PDA

View Full Version : Can someone post the Hollinger chat from today?


endrity
03-05-2007, 05:33 PM
Thanks to anyone that does.

jthig32
03-05-2007, 05:39 PM
Welcome to The Show! On Monday, NBA columnist John Hollinger will stop by to talk about the NBA season.
Send your questions now and join John in The Show Monday at 2 p.m. ET.
Hollinger archives: Columns (http://search.espn.go.com/keyword/search?searchString=john_hollinger) | Chats (http://sports.espn.go.com/chat/sportsnation/story?page=ChatArchiveHollinger)

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: Greetings everyone. For those who haven't seen it, Forbes magazine rated Kevin McHale as the No. 1 GM in sports this week. I can't wait for their upcoming issues that rate Darfur as the world's No. 1 tourist destination, Cheez Wiz as America's top food product, and "Cabin Boy" as the century's best film. On to the chat...

Firpo (North Arlington, NJ): I ask this constantly on these sites and no one ever answers me...If Steve Nash's career ended right now..is he a Hall of Famer? Keep in mind career averages of 14, 7.5, and 3, with no Finals appearrances.

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: It would be completely unprecedented for a player of Nash's quality not to make it. The career averages are meaningless because he took so long to get his career going.

Marcos (Miami,Fl.): Hey John, how do you feel on Wade's return and would he be healthy enough to guide Miami to repeat?

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: Here's the thing: Even if he comes back, I believe he can't return until there's only a week or two left in the season. Which means the supporting cast still has about a month to keep the ship afloat before we can even start talking about Miami's playoff scenarios. And I do worry about what's going to happen once he takes the first hit on that shoulder, especially since his whole game is driving to the basket and he takes about four nasty spills a game.

Mad Mike (Houston): John: Your take on the end of the Warrior/Wiz game yesterday? Also, what can the Warriors do to fix this team as it seems they have hit a wall. Thanks, MM

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: I don't know what Nelly's history is with those refs, and obviously I wasn't courtside to hear what was said. But as mad as the W's are about the call, if Nelly really dropped on F-bomb on the ref like then he should only be blaming himself for letting his temper cost the team a game. As far as Golden State's fixes, getting Baron Davis and Jason Richardson on the court at the same time would be a good start.

Matt (Glendale, CA): John, do you think the sixers will have two lottery picks, there own, and Denver's?

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: No, Denver will still make the playoffs, I don't think the Clips and Hornets are good enough to keep up with them. But that would be quite a coup for Philly if it turned out the other way, huh? And I don't believe the pick is protected.

Mad Mike (Houston): John: Baron will always have injuries. Should they limit him to 22-25 minutes a game and get a solid # 2 PG, to back him up? Thanks, MM

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: I'm really not sure that would prevent the injuries. In which case you'd be limiting his minutes for no reason whatsoever.

A.J. (Chicago): When Phoenix loses to lowly Philadelphia with Marion out of the lineup, where are all the columns proclaiming him the MVP?

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: This is a great point. If you looked at on-court vs. off-court stats last season, it was Marion, not Nash, that was the indispensable Sun; in fact it wasn't close. This year Nash's differential is better, but Marion's is still huge. It's just he misses so few games that people don't start on the "They're nothing without Marion" arguments the way they've done during Nash's absences.


Bill (Philly): Why are the Sixers winning and how can we make them stop?

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: For one thing, Andre Iguodala has played a whole lot better since around the Iverson trade. Some will pin this on Iverson, obviously, but even in the games AI missed the other AI wasn't putting up the kind of numbers he is now. As far as the lottery stuff, goes, don't sweat it, you're really not hurting your chances that much.

Tyler (nyc): Knicks have been playing with alot of heart, any chance they win the atlantic division this year?

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: Atnatic Division is a major stretch, even with Toronto struggling right now making up 4.5 games is asking a lot. Also, remember that the Nets will have Jefferson back sometime in the next week or so, and Knicks still trail them by half a game. That said, the miraculous recovery of Steve Francis (chortle, chortle) has helped, they only have two and a half games to make up on Miami and I believe they own the tie break.

Dennis, Honolulu: What if Oden and Durant look at the teams with the top 2 picks and decide to stay another year? Where does that leave Ainge, King, and all the other GM's banking on those youngsters saving their jobs?

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: I don't think they'd based their decision on that, just because it would basically cost them about $10 million five years down the road when they're on the last year of their rookie contract instead of the first year of a max extension. If they want to stay, they'll stay,but if they feel like turning pro, I don't think an internal "Memphis vs. Philadelphia" debate is likely to swing it.

Chris Elliot : THESE PIPES ARE CLEEEEEEAAAAAN!!!!

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: Wrong movie. Give me a Cabin Boy quote and then I'll be impressed. I saw it in the theater and, let me tell ya, that was about as bad a use of two hours as I've ever come up with.

jim (philly): Why would Oden stay in college when his teammates refuse to pass him the ball?

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: LOL ... the college game is not made to be easy for bigs. Teams just sag, sag, sag and force the guards to make all the plays.

Jon (Newark, NJ): Realistically are the nets going to make the playoffs? if they do, is there any chance they can get out of the first round?

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: Yes, with Jefferson coming back, Wade out and the Knicks and Magic looking feeble, I expect Jersey to be there. But I don't expect them to stick around for long.

Jeff (West Orange, NJ): I hear the Darfur tourism board is upset with you. Now they won't allow you to visit the country. Regrets?

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: Guess I'll have to change my plane tickets. Good thing I had Haiti as my Plan B.

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: (Since we're on the topic I should probably mention that what's happening in Darfur is actually really, really horrible and the world's response has been something less than inspiring. But labeling Kevin McHale as the most successful GM in sports is about the most patently idiotic thing I've ever seen, so much so that I didn't even realize until just now that Billy King was only a couple spots behind him.)

Ragan (pa): I understand you don't think tanking works (only three of the worst teams in the recent history have gotten #1 pick, etc), but with a win last night the sixers moved from 3rd worst, into a tie for 6th worst, and the way they are playing right now it wouldn't be shocking if they ended up with the 8-10th pick or higher! Regardless of the chances of getting Oden or Durant, that's a big difference from #3 for a rebuilding team, is it not?

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: I agree that Philly's situation is a little unusual since there are four teams at 22-38 and they're about to pass the other three, and that there's another pack of three teams with 35 losses that they're also in danger of passing. That said, they don't have many cupcakes left on the schedule so this may be their high-water mark.

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: I agree that Philly's situation is a little unusual since there are four teams at 22-38 and they're about to pass the other three, and that there's another pack of three teams with 35 losses that they're also in danger of passing. That said, they don't have many cupcakes left on the schedule so this may be their high-water mark.

Matt Hanna (NYC): John, what exactly was their basis for putting McHale number one? Seriously, did they give any insight to the criteria they used for rating the GMs?

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: they compared the team's record to that of their predecessor, and gave some kind of bonus for keeping salaries down. Since the T'wolves were an expansion team when McHale got there, the bar was set real low. Same goes for King and the mess that preceded him in Philly, which was how they won the Iverson lottery in the first place.

Ted, SLC: John, I need a power rankings explanation please: Last Thursday, Dallas beat Cleveland, but the next day, the Mavs' rating had dropped from 108.856 to 108.368 and the Cavs' rating went up from 102.966 to 103.289. How does that happen?

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: I'll give you two explanations: 1) there's a three-point bonus for the road team and the Mavs won by three, so the computer basically saw it as a draw, and 2) If that happens, as one might expect, the lower-rated team will tend to move up and the higher-rated team will tend to move down.

Ed (Michigan): If Kevin McHale is #1 wouldn't that make Matt Millen #2?

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: Clearly, an oversight by the folks at Forbes. And how do you make a list like this and leave out the entire Babcock family?

Jim (Boston): John, what would it take for the Bulls to get over the hump in the Eastern Conference (assuming Pau and KG aren't about to be waived)? Is there something more to it than the shooters just getting hot at the right time?

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: Even now, I think they're as good as anybody in the conference. They need Nocioni to come back and it would help if the two rookies start to turn the corner (Thomas sure looks like he's getting there), but mostly with the Bulls it comes down to whether the jumpers are falling.

Chris (NYC): Has Tyson Chandler finally figured it all out? His play elevated when Chris Paul went out, then elevated AGAIN when Paul came back.

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: It's hard not to be impressed with the improvement he's shown this season; to the point that one has to wonder if the Bulls would have been better off keeping him and not signing Wallace.

Ted, SLC: Gotcha. As thanks for the explanation, this one's for you... Nathanial (Cabin Boy): "I overheard the captain saying it would be jolly-jack splendid to spend a fortnight or two in Hawaii" Kenny: "Yea... that sounds like the captain... you do good impressions."

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: Now does everyone see what they missed? Anyone out there jealous they didn't see it? Anyone?

rick(Toronto): John, how likely are the Raptors to a) win the Atlantic and b) win a playoff round ?

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: Extremely likely to win the Atlantic; don't overreact to the past two games. As far as winning a playoff round, I'd like to see who they get as an opponent first. If they can get No. 3 and play a No. 6 Indiana team I like their odds a lot. A 4/5 battle with Cleveland or Chicago, not so much.

Miles WYO: Jerry Sloan coach of the year?

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: I'd still put Mitchell, Van Gundy and Avery Johnson ahead of him, but he'll get votes, and he's going to be a sentimental choice since he's never won.

Seba (Fort Worth): I think your power ranking formula doesn't account for how much a team gets up to play another. Once the Mavs hit double digits on their win streak and everyone is mentioning 70 wins they have been getting eveyone's A-game. How can you really fit that into the equation?

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: Even if there were validity to this, which I greatly doubt, are you really saying the zero-time champion Mavs are the team everybody would be all geeked up to play against?

Jack (Toronto): Raps have looked lacklustre at best in the last week or so. How important is il Mago (Bargniani) to this team? Not a good performance against the Cavs without him in the lineup.

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: I think it's a stretch to blame this on Bargnani's absence considering a) they gave up 120 points and b) Bargnani can't guard my keyboard, meaning that he probably couldn't have done much to prevent the outcome.

Ryan (NJ): I'm sure your tired of hearing it, but how do ya put the Spurs in front of the Mavs?...I'm very pro-Spurs, and I wouldn't even put the Mavs anywhere else but number 1, numbers or no-numbers.

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: Basically, the diference is that Dallas is beating people, while San Antonio is beating the tar out of people. They beat Houston on the road, by 23, without Tony Parker. That was the second night of a back-to-back; the previous night they won by 24 against Orlando. In their past eight games they aolso beat New Jersey by 25, Seattle by 31, Denver by 15, and Torontoy by 16 after being up 26 at halftime (though that latter datum doesn't go into the rankings). So San Antonio's play of late has been the more impressive of the two teams.

Jason (Lewes, DE): How many questions or comments do you recieve per session? Is there any rhyme or reason to what questions you choose?

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: I usually get around 1,000 questions per session, sometimes more if it's an interesting time of year. I pick them based on a) which ones show up on my screen of most recent questions, and b) which one of those are actually interesting and/or sane.

Matt Hanna (NYC): John, Dallas is 50-5 in there last 55. Why can't you just admit that maybe you need to fine-tune your formula? Don't be so stubborn. You'd be a liar if you didn't admit that when you saw the Spurs rank ahead of the Mavs, you didn't check your numbers over and over again.

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: The 50-5 is only relevant if we think it means they'll go 50-5 in their next 55 gtames. But Dallas doesn't have the point differential of a team that good (or of a 70-win team, for that matter); usually you need to be around +10 points per game to win at that clip. Which suggests they're not going to keep up this torrid pace forever, certainly not if it means eking out every 5-or-less game, which are basically 50-50 propositions.

David (Baltimore, MD): Youre explanation of Dallas V San Antonio is flawed. You claim recent point spread, however on your power rankings the spurs margin of victory for L25% is 10.53 while the Mavs is 11.73. SO how can the spurs be ahead of the mavs?

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: That was the condensed version, just pointing out how well they've played of late. Spurs bigger edge on Dallas is the season stats -- they have a stronger scoring margin and, an underrated factor, have played five more road games.

Jake (Oregon): Are you still thinking the Jazz are more likely to lose the first round in the playoffs than win? This looks like a team that can beat LA and Houston.

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: Yao Ming comes back tonight, so you might need to revise the Houston comment. Wouldn't be shocked if Utah won a round, but they'll be the underdog for sure.

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: Folks, that's allt he time I have for now, but I'll be back at my regular time next week and we can talk more NBA then.

jthig32
03-05-2007, 05:45 PM
Seba (Fort Worth): I think your power ranking formula doesn't account for how much a team gets up to play another. Once the Mavs hit double digits on their win streak and everyone is mentioning 70 wins they have been getting eveyone's A-game. How can you really fit that into the equation?

http://espn-ak.starwave.com/i/sn2.gif John Hollinger: Even if there were validity to this, which I greatly doubt, are you really saying the zero-time champion Mavs are the team everybody would be all geeked up to play against?

I'm sorry, but if John doesn't think teams get up more for games against elite teams, especially teams that are on a roll like the Mavs, then he's never played or watched sports, period.

Hell just watch a couple Mavs games. If you don't think Cleveland and Orlando threw their very best at us last week, you are insane.

endrity
03-05-2007, 05:55 PM
Thank for posting this, and yes I agree that the Mavs will have everybody's best shot until the end of the season.

LRB
03-05-2007, 06:23 PM
Hollinger = over educated idiot

"The 50-5 is only relevant if we think it means they'll go 50-5 in their next 55 gtames. But Dallas doesn't have the point differential of a team that good (or of a 70-win team, for that matter); usually you need to be around +10 points per game to win at that clip. Which suggests they're not going to keep up this torrid pace forever, certainly not if it means eking out every 5-or-less game, which are basically 50-50 propositions. "

So does this mean if the mavs win 70+ games and have a point differential significantly below +10, that Hollinger's theories are self admitted full of shit and not worth the price of used toliet paper?

Dirkadirkastan
03-05-2007, 08:10 PM
So does this mean if the mavs win 70+ games and have a point differential significantly below +10, that Hollinger's theories are self admitted full of shit and not worth the price of used toliet paper?

Save this for a future Hollinger chat question. ;)

wmbwinn
03-05-2007, 09:10 PM
Ted, SLC: John, I need a power rankings explanation please: Last Thursday, Dallas beat Cleveland, but the next day, the Mavs' rating had dropped from 108.856 to 108.368 and the Cavs' rating went up from 102.966 to 103.289. How does that happen?

John Hollinger: I'll give you two explanations: 1) there's a three-point bonus for the road team and the Mavs won by three, so the computer basically saw it as a draw, and 2) If that happens, as one might expect, the lower-rated team will tend to move up and the higher-rated team will tend to move down.
------------------------------------------------------------

The above logic that the home team has a 3 point handicap is ludicrous. The Suns have only lost one game on the road in the Eastern conference. The Mavs seemed to play better on the road in the playoffs last year than they did at home. The road/home issue is over-rated. More evidence of a flawed program/formula

Male30Dan
03-05-2007, 09:13 PM
Any team ranked ahead of the Mavs right now is just rediculous!

wmbwinn
03-05-2007, 09:14 PM
Ryan (NJ): I'm sure your tired of hearing it, but how do ya put the Spurs in front of the Mavs?...I'm very pro-Spurs, and I wouldn't even put the Mavs anywhere else but number 1, numbers or no-numbers.

John Hollinger: Basically, the diference is that Dallas is beating people, while San Antonio is beating the tar out of people. They beat Houston on the road, by 23, without Tony Parker. That was the second night of a back-to-back; the previous night they won by 24 against Orlando. In their past eight games they aolso beat New Jersey by 25, Seattle by 31, Denver by 15, and Torontoy by 16 after being up 26 at halftime (though that latter datum doesn't go into the rankings). So San Antonio's play of late has been the more impressive of the two teams.
--------------------------------------------------------------------------

go read axelamenos (I hope I spelled that right, going off memory) post where he argues that the Hollinger system is based on an arbitrary assumption that the last 25% of the season is the range of games to address for "recent" numbers. That well written post demonstrated that if the percentage were the last 35% of the season or the last (most recent) 15% of the season, that the Mavs would be #1.

The Hollinger formula is based on an arbitrary assumption or choice of the unqualified/undefended use of a rigid 25% of the most recent games...

wmbwinn
03-05-2007, 09:24 PM
Matt Hanna (NYC): John, Dallas is 50-5 in there last 55. Why can't you just admit that maybe you need to fine-tune your formula? Don't be so stubborn. You'd be a liar if you didn't admit that when you saw the Spurs rank ahead of the Mavs, you didn't check your numbers over and over again.

John Hollinger: The 50-5 is only relevant if we think it means they'll go 50-5 in their next 55 gtames. But Dallas doesn't have the point differential of a team that good (or of a 70-win team, for that matter); usually you need to be around +10 points per game to win at that clip. Which suggests they're not going to keep up this torrid pace forever, certainly not if it means eking out every 5-or-less game, which are basically 50-50 propositions.

David (Baltimore, MD): Youre explanation of Dallas V San Antonio is flawed. You claim recent point spread, however on your power rankings the spurs margin of victory for L25% is 10.53 while the Mavs is 11.73. SO how can the spurs be ahead of the mavs?

John Hollinger: That was the condensed version, just pointing out how well they've played of late. Spurs bigger edge on Dallas is the season stats -- they have a stronger scoring margin and, an underrated factor, have played five more road games.
---------------------------------------------------------------------
http://www.dougstats.com/06-07Teams.html
1) the Mavs at this point have averaged 8.6 points per game scored more than they have allowed to be scored against them. The Spurs are a close 8.5 for the season. So, the Spurs are not superior for the season in this category.

2)The Mavs are the ONLY team in the NBA ranked in the top ten in both the offensive category AND the defensive category (based on points scored for offense and points allowed for defense):

Team Scoring Team Defense
Team Pts/Game Team Pts/Game
1 PhoenixSuns 110.5 1 SanAntonioSpurs 90.0
2 DenverNuggets 105.0 2 HoustonRockets 91.1
3 GSWarriors 104.9 3 DallasMavericks 91.8
4 WashingtonWizards 104.8 4 DetroitPistons 92.2
5 LALakers 103.3 5 ClevelandCavaliers 93.4
6 UtahJazz 102.6 6 OrlandoMagic 94.2
7 SacramentoKings 101.0 7 ChicagoBulls 94.9
8 MemphisGrizzlies 100.9 8 NOrleansHornets 95.2
9 DallasMavericks 100.4 9 LAClippers 96.4
10 MilwaukeeBucks 100.4 10 MiamiHeat 97.7

3)At some point, he will have to figure out that the hypothesis that winning games where the score is within a 5 point margin is a 50/50 proposition is a flawed analysis. I've forgotten the poster's name who did the math on the probabilities on the Mavs' victories in close games that shows that the chance that the Mavs could win so many close games on a probability of 50/50 is ludicrous. It was statistically determined that the odds for the Mavs were around 85/15 in favor of a Mavs' win in close games. Anyway, that great post on this board should be referenced and given credit. Hollinger is wrong here also.

wmbwinn
03-05-2007, 09:35 PM
Ryan (NJ): I'm sure your tired of hearing it, but how do ya put the Spurs in front of the Mavs?...I'm very pro-Spurs, and I wouldn't even put the Mavs anywhere else but number 1, numbers or no-numbers.

John Hollinger: Basically, the diference is that Dallas is beating people, while San Antonio is beating the tar out of people. They beat Houston on the road, by 23, without Tony Parker. That was the second night of a back-to-back; the previous night they won by 24 against Orlando. In their past eight games they aolso beat New Jersey by 25, Seattle by 31, Denver by 15, and Torontoy by 16 after being up 26 at halftime (though that latter datum doesn't go into the rankings). So San Antonio's play of late has been the more impressive of the two teams.
--------------------------------------------------------------------------

go read axelamenos (I hope I spelled that right, going off memory) post where he argues that the Hollinger system is based on an arbitrary assumption that the last 25% of the season is the range of games to address for "recent" numbers. That well written post demonstrated that if the percentage were the last 35% of the season or the last (most recent) 15% of the season, that the Mavs would be #1.

The Hollinger formula is based on an arbitrary assumption or choice of the unqualified/undefended use of a rigid 25% of the most recent games...

http://www.dallas-mavs.com/vb/showthread.php?t=29008

go read Axelamenos's post there...

wmbwinn
03-05-2007, 09:57 PM
http://www.dallas-mavs.com/vb/showthread.php?t=28982&highlight=50%2F50+proposition

explains the value of using a points differential or margin of victory ranking...

--------------------------------------------


"From Fish at Dallas Basketball:
If point differential is the be-all and end-all, then the Points Differential Champion should win the NBA Finals, right? Or at least that team should make it to the Finals, right?

Here are the last four years of Finals participants, and their point differential rank:

2006 Miami (5th) over Dallas (3rd)

2005 SA (4th) over Det (5th)

2004 Det (2nd) over LAL (7th)

2003 SA (3rd) over NJ (4th)

Now, I'm no statgeek. But I can read. And I don't see the team with the top differential there anywhere."

-------------
nice post, left texas

wmbwinn
03-05-2007, 10:05 PM
Here are the last 15 NBA champs, along with their league wide ranking in win percentage and point differential:

Year Champ Win % Point Differential
2006 Heat 5th 5th
2005 Spurs 2nd 1st
2004 Pistons 6th 2nd
2003 Spurs 1st 3rd
2002 Lakers 2nd 2nd
2001 Lakers 2nd 7th
2000 Lakers 1st 1st
1999 Spurs 1st 1st
1998 Bulls 1st 1st
1997 Bulls 1st 1st
1996 Bulls 1st 1st
1995 Rockets 10th 10th
1994 Rockets 2nd 6th
1993 Bulls 3rd 2nd
1992 Bulls 1st 1st

Average rank: win %: 2.6, point diff.: 2.9

Over that time, winning percentage has an edge over point differential in predicting the champion, but its very slight.

http://www.dallas-mavs.com/vb/showthread.php?t=28446&page=2&highlight=probability

post #54

Another nice history exam that shows that winning percentage is a better predictor thatn point differential...

that refutes Hollinger squarely....

wmbwinn
03-05-2007, 10:14 PM
It took forever, but I finally found the post that shows that Hollinger's theory that games decided by 5 points or less are a 50/50 proposition is false for the Mavs...
--------------------------------------------

"Let's look at this from the other angle.

The Mavs are 21-4 in games decided by 9 or fewer points. Is that really determined by "luck"?

If close games were a matter of a coin flip, i.e. the probability the Mavs pull out a win in a close game is 50%, then the odds of winning 21 or more out of 25 is 0.04%. So it probably isn't pure luck.

What if they were 60% likely to pull out a single close win? Still only a 0.95% shot at winning 21+ of 25.

70%? A modest 9.05%.

At 80% there is finally a decent chance with 42.07%. For there to be a 50% chance of winning 21+ out of 25, there would have to be an 81.57% chance of winning a single close game.

Remarkably, you get almost the exact same result for the Mavs' 12-2 record in games decided by 5 or less. In order for the odds of winning 12+ out of 14 to be 50%, the odds for any one game must be 81.35%.

Yes, I think winning close games is a matter of skill, not luck"

http://www.dallas-mavs.com/vb/showthread.php?t=28446&page=4&highlight=probability

post 122

wmbwinn
03-05-2007, 10:36 PM
++++++++++++++++++++++++++++++++++
http://www.courtsidetimes.net/articles/337/
by Justin Kubatko | permalink | trackback | comments |
Visit the author's site: http://www.basketball-reference.com


MFFL quoted the above article, see
http://www.dallas-mavs.com/vb/showthread.php?t=28446&page=3&highlight=probability

post 118

---------------------------------

Basically, this article concludes that winning close games is something that 27% of NBA champs have failed to do, but that winning frequent blowouts is something most champs do.

So, I looked at Febuary and March to date for the Suns, Mavs, and Spurs:

Suns: won 3 blowouts and were blown out once (by the way, the quoted article defines a blowout as a 15 point difference in a game).

Spurs: won 7 blowouts and lost by a blowout margin twice

Mavs: won 6 blowouts and haven't lost at all in Febuary and March... much less losing by a blowout margin.

Actually, the Mavs have only lost 3 games by a blowout margin. They were losses to the Rockets on Nov. 4, Clippers Nov. 8, and the Jazz Dec. 11

The Spurs have only had 3 blow out losses but two of them happened in Febuary and Hollinger is saying that the Spurs are great based on recent games...

The Suns only have one blowout loss for the season and it also happened in Febuary.

wmbwinn
03-05-2007, 10:37 PM
Hollinger annoys me. I'll have to get over it and quit dwelling on it I suppose. I hope quoting other posters isn't against the rules...

rmacomic
03-05-2007, 10:44 PM
[QUOTE= Chris Elliot : THESE PIPES ARE CLEEEEEEAAAAAN!!!!

John Hollinger: Wrong movie. Give me a Cabin Boy quote and then I'll be impressed. I saw it in the theater and, let me tell ya, that was about as bad a use of two hours as I've ever come up with.

This line was in the movie, in fact it was even used in the trailer. If you actualy had seen the film you might now. Also you would have seen some of the best inane chatter comedy writing of the 90's. Lines like. "I'm sorry I wasn't listening to you. I was just think about how much fun it is to roast pumpkin seeds." Even David Letterman in one of his few movie roles.

Asshole probably just used Cabin Boy because his editor won't let him reference Ishtar anymore.

Ted, SLC: John, I need a power rankings explanation please: Last Thursday, Dallas beat Cleveland, but the next day, the Mavs' rating had dropped from 108.856 to 108.368 and the Cavs' rating went up from 102.966 to 103.289. How does that happen?

John Hollinger: I'll give you two explanations: 1) there's a three-point bonus for the road team and the Mavs won by three, so the computer basically saw it as a draw, and 2) If that happens, as one might expect, the lower-rated team will tend to move up and the higher-rated team will tend to move down.

3) Also, I don't really have a system. Here at ESPN we do lowest common denominator reporting, if we don't mention enough recognizable names in a week we don't fill our quota. So putting The Spurs ahead of the Mavs guarantees the morons who actually value what we say, will keep coming back. It's the same reason guy's like Ron Artest and Allen Iverson grace the cover of our magazine, even though they haven't been relevant all year. Sports Sensationalism. Dumbing down the American sports fan.

Seba (Fort Worth): I think your power ranking formula doesn't account for how much a team gets up to play another. Once the Mavs hit double digits on their win streak and everyone is mentioning 70 wins they have been getting eveyone's A-game. How can you really fit that into the equation?

John Hollinger: Even if there were validity to this, which I greatly doubt, are you really saying the zero-time champion Mavs are the team everybody would be all geeked up to play against?

Anyone want to sniff glue?

Ryan (NJ): I'm sure your tired of hearing it, but how do ya put the Spurs in front of the Mavs?...I'm very pro-Spurs, and I wouldn't even put the Mavs anywhere else but number 1, numbers or no-numbers.

John Hollinger: Basically, the diference is that Dallas is beating people, while San Antonio is beating the tar out of people. They beat Houston on the road, by 23, without Tony Parker. That was the second night of a back-to-back; the previous night they won by 24 against Orlando. In their past eight games they aolso beat New Jersey by 25, Seattle by 31, Denver by 15, and Torontoy by 16 after being up 26 at halftime (though that latter datum doesn't go into the rankings). So San Antonio's play of late has been the more impressive of the two teams.

Of course this argument was never near my lips when I had the Suns ranked above the Mavs after they won 17 in a row. Have you seen Nash pass? Wow, it's like watching John Stockon... if he played defense like a girl who plays defense like a gay.

Matt Hanna (NYC): John, Dallas is 50-5 in there last 55. Why can't you just admit that maybe you need to fine-tune your formula? Don't be so stubborn. You'd be a liar if you didn't admit that when you saw the Spurs rank ahead of the Mavs, you didn't check your numbers over and over again.

John Hollinger: The 50-5 is only relevant if we think it means they'll go 50-5 in their next 55 gtames. But Dallas doesn't have the point differential of a team that good (or of a 70-win team, for that matter); usually you need to be around +10 points per game to win at that clip. Which suggests they're not going to keep up this torrid pace forever, certainly not if it means eking out every 5-or-less game, which are basically 50-50 propositions.

55 games or 67% of a season is not enough to go by, however the last 25 games or 30% is a much better judge of how a team will do. That's what we call conveinent logic. If it makes what you're already going to say work, use it instead.

David (Baltimore, MD): Youre explanation of Dallas V San Antonio is flawed. You claim recent point spread, however on your power rankings the spurs margin of victory for L25% is 10.53 while the Mavs is 11.73. SO how can the spurs be ahead of the mavs?

John Hollinger: That was the condensed version, just pointing out how well they've played of late. Spurs bigger edge on Dallas is the season stats -- they have a stronger scoring margin and, an underrated factor, have played five more road games.

The expanded version also includes my invite to Tony and Eva's wedding. Isn't it great how I am constantly changing the numbers I use, Last 25 games or the entire season, or whatever keeps you guessing. When they said insanity is the best defense they weren't kidding. It's hard to fire a writer when he's sniffing your crotch and howling at the moon. Huntr Thompson eat your heart out.

QUOTE]

Dirkadirkastan
03-06-2007, 02:21 AM
I remember that article posted by MFFL, and I thought it was very strange.

It compares close game analysis versus blowout analysis, and concludes that while close games do not make good predictions for who will be the champion, champions do tend to win most blowouts.

In my opinion, it's a leap to believe that blowouts can be analyzed the same way as close games. That's because of the different nature of the two. A close win can easily turn into a close loss in the final minutes of a game and vice versa, but obviously that's not true for blowouts, i.e. a blowout win will not suddenly turn into a blowout loss, at least not in the same fashion. The relation between blowout wins and losses is not the same as the relation between close wins and losses.

It's no surprise elite teams blow out their opponent more often than they themselves get blown out, but I don't really see what blowout wins and blowout losses have to do with each other in the first place. To me, they're on opposite ends of the spectrum: blowout wins, close wins, close losses, blowout losses. I would be interested in seeing an explanation why blowout wins are significant and close wins are not. In other words, it seems contradictory to assume there's no skill at preventing a close win to becoming a close loss, while it's praiseworthy to prevent a blowout win from becoming a supposedly insignificant close win.

______________________

It's difficult to review the entire NBA to determine whether winning close games is a matter of luck or skill. First of all, it's a zero sum rule, so even if two teams playing head-to-head don't know what they're doing down the stretch, somebody is going to pull out the win anyway. So for those teams, winning close games is a matter of luck, and some will be lucky and have a good close game record. But that doesn't mean it's a matter of luck for everybody. After all, just about anyone on this forum can make an NBA 3-pointer, and for most of us, it's a matter of luck. But for sharpshooters in the NBA, it's clearly not.

Furthermore, no one ever claimed winning close games was a required skill for winning a championship, just that it was a skill. And just like any other skill, different elite teams will possess different degrees of that skill. So if there are a few champions in history that have a bad track record in close games, that doesn't necessarily mean nobody was ever good at winning close games; it probably just means those teams made up for that deficiency in some other department.

Usually Lurkin
03-06-2007, 06:23 AM
dp

Usually Lurkin
03-06-2007, 06:32 AM
It took forever, but I finally found the post that shows that Hollinger's theory that games decided by 5 points or less are a 50/50 proposition is false for the Mavs...
--------------------------------------------

"Let's look at this from the other angle.

The Mavs are 21-4 in games decided by 9 or fewer points. Is that really determined by "luck"?

If close games were a matter of a coin flip, i.e. the probability the Mavs pull out a win in a close game is 50%, then the odds of winning 21 or more out of 25 is 0.04%. So it probably isn't pure luck.


I don't think the mavs w/l record is a fluke of chance, but . . .
Dirkadirkastan's coin flipper says that a 21 and 5 record should happen 4 out of every 100 26-game sets that are decided by 9 or less points, right? That's 1 out of every 25 times. There are 30 teams in the NBA. How many have 26 games decided by 9 or less points? In the history of the NBA, how many teams have had 26 games decided by 9 or less points? If Hollinger is correct, and those games are decided by chance, then 21 wins out of 26 will eventually happen.


go read axelamenos (I hope I spelled that right, going off memory) post where he argues that the Hollinger system is based on an arbitrary assumption that the last 25% of the season is the range of games to address for "recent" numbers. That well written post demonstrated that if the percentage were the last 35% of the season or the last (most recent) 15% of the season, that the Mavs would be #1.

The Hollinger formula is based on an arbitrary assumption or choice of the unqualified/undefended use of a rigid 25% of the most recent games...
That's not how it works. His numbers were (probably) chosen under some criteria. Most likely he chose the numbers that historically have best predicted who would win the next game, or who would win the playoffs, or something.



"From Fish at Dallas Basketball:
...Here are the last four years of Finals participants, and their point differential rank:

2006 Miami (5th) over Dallas (3rd)
2005 SA (4th) over Det (5th)
2004 Det (2nd) over LAL (7th)
2003 SA (3rd) over NJ (4th)

Now, I'm no statgeek. But I can read. And I don't see the team with the top differential there anywhere."

Those numbers are wrong. I think Fish corrected this in an update. SA was 1st, I believe. We barely beat them.

Dirkadirkastan
03-06-2007, 06:48 AM
I don't think the mavs w/l record is a fluke of chance, but . . .
Dirkadirkastan's coin flipper says that a 21 and 5 record should happen 4 out of every 100 26-game sets that are decided by 9 or less points, right? That's 1 out of every 25 times. There are 30 teams in the NBA. How many have 26 games decided by 9 or less points? In the history of the NBA, how many teams have had 26 games decided by 9 or less points? If Hollinger is correct, and those games are decided by chance, then 21 wins out of 26 will eventually happen.



POINT 04%. That's 1 in 2500!

Of course, now that they're 23-4, the updated number is 0.0155% (1 in 6400)

Usually Lurkin
03-06-2007, 07:36 AM
POINT 04%. That's 1 in 2500!

Of course, now that they're 23-4, the updated number is 0.0155% (1 in 6400)
my bad. I read it wrong and didn't think it out. It would be interesting to count up the number of (now) 27 game stretches of close games in the history of the NBA and plot out win percentages. Though it's unlikely that .01% will occur for any given 27 game stretch, given enough stretches, it's likely that it will happen somewhere.

We had a discussion in a stats class that I think illustrates a limitation of stats. One of the profs plotted out the world record breaking performances in some sport (long jump or 100 yard dash, I think), and compared the improvement curve to what would be expected by chance, and determined that the same curve would result if you started out drawing randomly from a single population of results. Very few people think that improvement in world records occurs by chance. And those are probably the same people that might think the Mavs record has occurred by chance.

alexamenos
03-06-2007, 10:13 AM
John Hollinger: I'll give you two explanations: 1) there's a three-point bonus for the road team and the Mavs won by three, so the computer basically saw it as a draw...

I like how Hollinger says that "the Computer" saw it as a draw as if a computer is capable of watching a basketball game and reading the boxscores and then making some judgment regarding the outcome of the game.

Perhaps Hollinger was just more comfortable placing the blame on the computer rather than saying "I instructed the computer to view a Mavericks win over the Cavaliers as a Cavs win over the Mavs."

Usually Lurkin
03-06-2007, 11:38 AM
I like how Hollinger says that "the Computer" saw it as a draw as if a computer is capable of watching a basketball game and reading the boxscores and then making some judgment regarding the outcome of the game.

Perhaps Hollinger was just more comfortable placing the blame on the computer rather than saying "I instructed the computer to view a Mavericks win over the Cavaliers as a Cavs win over the Mavs."

That reminded me of an old NFL football tapes segment that had a lowly loser home team grinding out a 3 point loss to a better visiting team. They kicked a final second field goal to reduce the margin to 3 points. The home crowd went wild the losing players jumped for joy and dumped gatorade on their coach. "They beat the spread! They beat the spread! They beat the spread!" the announcer repeated.

LRB
03-06-2007, 01:04 PM
What Hollinger is all of his anal retentive wisdom fails to see is that the object of NBA games is winning, not beating the spread, not coming close to winning, but winning. And a win by 1 point counts the same as a win by 100 points.

Now the object of an NBA season is to win a championship or if that is not a realistic goal to get closer to winning a championship in the future. To do this may cause your team to win by a lesser margin or even lose less games. The mavs are focused on winning the championship to the point that they are well prepared to lose regular season games to get themselves ready. However they keep winning in spite of this. But unlike the Suns who play their short rotation deep into the 4th quarter even when leading by 20+, Dallas tries to play a much larger rotation to prepare lesser players for the chance they are needed in the playoffs and to save their key players by limiting their minutes. Hollingers dumbass formula has no comprehension of this. His formula doesn't take into account injuries, back 2 back games, teams getting up to play the hot team, big leads lost by garbage time players, teams just having hot nights or cold nights (after all the NBA is a league of runs), or a host of other variables.

Yet Hollinger in all of his arrogant pompus assness, tries to preach that his formula is right and infalable. He's a jerk, a dumbass, and a couldn't find his own ass in the dark with a flashlight when it comes to his precious little formula which is based on junk statistics.

chumdawg
03-06-2007, 10:08 PM
Tonight's game is the kind of thing Hollinger is talking about. With a couple minutes to go, it looked like it could easily come down to a one-possession game. But the Mavs pulled away and put it in the books by 13.

A lesser team is more likely to let it come down to the last couple possessions. This is why win margin does an excellent job, for the most part, of describing a team's performance over the course of the season.

wmbwinn
03-06-2007, 10:18 PM
As posted above, if Hollinger is right, then a game decided by five points or less is a 50/50 proposition. If he right, the Mavs should have as many losses as wins in games decided by less than five points...

But, Hollinger is not right. If Hollinger is right and the Mavs' season is a fluke of statistical probabilities, then the Mavs are playing 1:6400 odds.
That is ludicrous.

It is silly to take one stat like points differential and try to turn it into a tool to predict the NBA game.

Now, if you go back 8 years or more, then Hollinger's model did seem to predict the outcome. But, the game has changed with rule changes and the evolution of the calls/refs and perhaps other factors. In the last 8 years or so, Hollinger's model has not predicted anything accurately.

There are other factors to consider. Hollinger (or a better statistics man) needs to discover why the Mavs are playing 1:6400 odds successfully and winning almost all of their games when the difference is five points or less. There has to be a modifying discovery that will change the formula to account for this anomaly (it is not an anomaly, it is evidence that Hollinger's formula is inaccurate in the NBA of the last almost decade).

I personally think that the shots taken differential is a big part of that modification. I have discussed this at length in the past.

Dirkadirkastan
03-06-2007, 10:25 PM
I agree with your post, wmb, but for clarity, I'll just say that the 1/6400 is actually for nine points or less. That's because I did my own work separately from Hollinger and didn't catch his definition of "close game" at five points.

But for five points or less, it's still pretty far out there. At 50%, the likelihood of going 14-2 is around 1/478.

wmbwinn
03-06-2007, 10:28 PM
I agree with your post, wmb, but for clarity, I'll just say that the 1/6400 is actually for nine points or less. That's because I did my own work separately from Hollinger and didn't catch his definition of "close game" at five points.

But for five points or less, it's still pretty far out there. At 50%, the likelihood of going 14-2 is around 1/478.

Thanks for the clarification. It is still crazy to think that 1:478 odds can be explained away by anything that Hollinger has ever offered as an explanation.

Windmill360
03-06-2007, 10:34 PM
Hollinger should make a stat that says the difficulty of the game due to the size of the target on our back...... with every win it increases.

I don't think any other team friggin plays as hard (relatively) against anyone other than us

Did you ever think of that, Hollinger, you ass?

wmbwinn
03-06-2007, 10:39 PM
Hollinger should make a stat that says the difficulty of the game due to the size of the target on our back...... with every win it increases.

I don't think any other team friggin plays as hard (relatively) against anyone other than us

Did you ever think of that, Hollinger, you ass?

Agreed. The last 7 many games have felt like the opposing team was playing a playoff game against us. And, we have repeatedly knocked them out in the fourth quarter...

LRB
03-06-2007, 11:20 PM
Tonight's game is the kind of thing Hollinger is talking about. With a couple minutes to go, it looked like it could easily come down to a one-possession game. But the Mavs pulled away and put it in the books by 13.

A lesser team is more likely to let it come down to the last couple possessions. This is why win margin does an excellent job, for the most part, of describing a team's performance over the course of the season.

Chum if Hollinger presented is arugment in a more resonable fashion such as more times than not or most of the time instead it's always true, I wouldn't give him such a hard time. I can see the argument that generally, teams with a higher MOV are playing at a higher level, as long as you allow for exceptions. Hollinger doesn't seem to want to allow for exceptions.

for example Hollinger states "John Hollinger: The 50-5 is only relevant if we think it means they'll go 50-5 in their next 55 gtames. But Dallas doesn't have the point differential of a team that good (or of a 70-win team, for that matter); usually you need to be around +10 points per game to win at that clip. Which suggests they're not going to keep up this torrid pace forever, certainly not if it means eking out every 5-or-less game, which are basically 50-50 propositions. " However he ignores that over the last 15 games Dallas is winning at an average of +10 points per game. And 16 > 25% of the 60 games that the Mavs have played by 1 game. Remember how Hollinger worships at the statistical alter of how a team performs over the last 25% of games played? He contradicts himself, projects his statistical analysis as absolute maximums without exception when at best they are just general rules of thumb. Surely he knows enough or should know enough to know he's wrong. So he's either a lying ahole or he's a really stupid dumbass or possibly both.

wmbwinn
03-06-2007, 11:31 PM
The Spurs just beat Portland by smashing them in the fourth quarter 34-20. The Spurs hit four consecutive 3 pointers to kill Portland. Portland had the Spurs on the ropes all night and appeared poised to end the Spurs' winning streak.

That 14 point win in the fourth quarter meant the Spurs won by five points.

According to Hollinger, the Spurs won a 50-50 coin flip because "games decided by five points or less are a 50/50 proposition".

The Spurs did not win by luck or by the chance of a coin flip. They put the clamps on Portland in the last 3 minutes of the game and put their 3 best shooters into the game.

See some of my earlier posts about the Spurs in other threads. The Spurs are a very successful 3 point shooting team who defends against the 3 point shot very well and is ranked #2 on total defense. The Spurs are ranked #2 in FG% differential (they shoot well and prevent the other team from shooting well). The Spurs just did what the Spurs do best when it mattered. And, they did not give up when they were down 7 points with 3 minutes left in the game after they had been behind about the same amount the whole game...

Hollinger doesn't get it.

wmbwinn
03-06-2007, 11:38 PM
Go look at the boxscore on the Portland/Spurs game. Portland beat the Spurs in every single statistical category EXCEPT for 3 pointers made. And, the difference in the number of 3 pointers made were those 4 made consecutively at the end of the game. Portland has to be sick to their stomachs...

chumdawg
03-07-2007, 12:31 AM
A few things...

What Hollinger has accomplished--quite shrewdly, I might add--is to put basketball statwork "on the map," so to speak, in a similar manner that the baseball guys did a few years in back. That in itself is an accomplishment, and it's not one that I mind seeing. The fact that we all are talking about goes to show what a good job he/ESPN did.

Now, I highly suspect that the ESPN part of the "Hollinger/ESPN" subject had more to do with it than anything else. Hollinger is a very, very sharp stats guy, and I suspect that he knows far better than the rest of us what limitations his model has. But he was given, by ESPN, the job of designing a "power ranking" measure and he gave it his best effort. I *do* believe that his formula is his best effort at that task. He might take a little heat based on the outliers at the top side, but don't you know that ESPN is loving it. I mean, how many fans heard Hollinger's name for the first time this season?

So, we should really view it all for what it is. Namely, a design to get us all talking basketball. Bravo, in that regard.

Now, about point differentials. They hold up VERY well over the course of a lot of NBA history. They have usefulness besides just the team level. They are used on a unit level and all the way down to the individual player level. Again, they are VERY useful stats. And they are useful because they can be shown with statistical certainty to correlate very strongly with win percentages. Frankly, for anyone to argue otherwise would be entirely counterintuitive and, to me at least, borderline delusional.

But still, you must recognize what they are. Pythagorean win formulas (meaning, basically, the value of the exponent used) are designed in such a way that they will correlate most strongly to the large body of statistical history we have to work with. In short, they are correct because they are the most correct model we can find to match the historical data.

But they are still known to have some trouble on the edges. In other words, they do a whole lot better job of telling you that a team with a 0.00 margin of victory will have a .500 record than telling you that a team with a +10.00 margin of victory will win 70 games. And this is because of one good reason--one reason we are seeing this year with our Mavs: better teams do a better job of winning close games. So sometimes you will get a situation where a team has more wins than their margin of victory (or Pythag, or what have you) would suggest they should.

But don't throw the baby out with the bathwater. Consider, for example, two teams that both play the same five average opponents. One team wins all five games, by one point each. The other wins four games by 20 points apiece and loses the fifth game by one point. Do you believe the 5-0 team is demonstrably better than the 4-1 team?

The whole reason for the MOV stats is to look deeper than win percentage. If you simply want to know who won at the highest rate, you don't need to look any deeper than the results in the paper. Win percentage is already tracked. Stats guys are trying to get deeper. And sometimes they chance upon some interesting things. As I mentioned upthread, tonight's game is a great example of one team's exerting its superiority over the other. The Mavs/Nets game was within two possessions with two minutes left, and it could have easily devolved into a last-shot scenario. But the Mavs exerted their superiority before that happened. Shouldn't that be measurable by some parameter? Shouldn't we know that the Mavs beat the Nets by more than just a buzzer-beater that went in?

Consider a scenario where the Mavs play the Celtics 82 times in a year, but it's a one-minute game. The Celtics are going to win a lot of those contests. Stretch it out to a full 48 minutes, and the Celtics are going to be hard pressed to win more than a handful. This is what Hollinger is getting at when he talks about close games going either way. They do, more often than it seems some of you would like to believe. But the teams that don't let the games get down to one possession are, demonstrably, better teams.

Take the limitations in Hollinger's formula for what they are (and what they are is primarily a vehicle for media attention rather than inscrutably honest stats work), but don't ignore the underlying methodology. Hollinger is, most certainly, not a moron.

Five-ofan
03-07-2007, 12:35 AM
am i the only one that finds hollingers ripping of forbes funny? he rips forbes for coming up with a statistical formula and stubbornly sticking by it no matter how obvious it is to the human eye that the result is wrong... sound like anyone else we know??

Dirkadirkastan
03-07-2007, 12:48 AM
it seems contradictory to assume there's no skill at preventing a close win to becoming a close loss, while it's praiseworthy to prevent a blowout win from becoming a supposedly insignificant close win.

This is the biggest flaw in the coin-flip analysis, in my opinion.

chumdawg
03-07-2007, 12:55 AM
I don't understand. Are you saying that the team ahead by ten in the closing minutes that holds on to win by ten is not a more accomplished team than the one ahead two in the closing minutes that holds on to win by two?

What are you saying?

Dirkadirkastan
03-07-2007, 01:12 AM
According to Hollinger, if you let a game slip from, say, a 12 point lead to a 4 point win, then you've lost all accomplishment and left the game to a coin flip, but if you let a 4 point lead slip to a 4 point loss, that's OK because the game was already in the realm of a coin flip anyway.

Personally, I think the latter is far worse than the former, but Hollinger seems to claim the opposite.

alexamenos
03-07-2007, 10:18 AM
So, we should really view it all for what it is.

according to stein, his power-rankings were the most frequently viewed page on the espn website. espn desired to get more of this, and hence they desired to have a power-ranking which would be updated daily...enter Hollinger.

chumdawg
03-07-2007, 10:47 AM
According to Hollinger, if you let a game slip from, say, a 12 point lead to a 4 point win, then you've lost all accomplishment and left the game to a coin flip, but if you let a 4 point lead slip to a 4 point loss, that's OK because the game was already in the realm of a coin flip anyway.

Personally, I think the latter is far worse than the former, but Hollinger seems to claim the opposite.No, he doesn't claim that. In fact, he's on board with you. The first scenario gives the team a plus-four margin while the second scenario gives the team a minus-four margin. That's an 8-point swing in collective margin of victory. As you can see with where the Mavs and Spurs are at this season, an 8-point swing goes a long, long way (if you average that each night).

So, what your example does is justify exactly what Hollinger is trying to do. The team that goes from 12 up to 4 up at the buzzer is, according to his model, a *much* better team than the one that goes from 4 up to 4 down. As a full-season model, anyway.

The "close game" argument can be looked at by itself, but in Hollinger's model it all comes out in the wash by the end.

Flacolaco
03-07-2007, 10:50 AM
Has he publicly applied his model retro-actively to past seasons playoff-successful teams to validate this "best predictor of success" claim?

alexamenos
03-07-2007, 11:31 AM
It is silly to take one stat like points differential and try to turn it into a tool to predict the NBA game.

It is silly, and if ESPN/Hollinger were more interested in statistical models than website hits they'd acknowledge as much.

To repeat myself yet again and again...

models such as these do a lousy job of telling you whether the mavs will beat the suns in a 7 game series, but they do a great job of telling you that the Spurs would beat the Charlotte Bobcats in 7 game series. That is...

--they're great for telling you things which you already know, but don't care about...
--they're lousy for telling you things which you don't know, but do care about....

Hollinger (or a better statistics man) needs to discover why the Mavs are playing 1:6400 odds successfully and winning almost all of their games when the difference is five points or less.

color me entirely unconvinced that winning close games is luck. That is, I am ignorant of any compelling argument or statistical evidence that this is in fact true. anyhoo....

chumdawg above makes a good illustration when he notes that the Celtics would beat the Dallas quite frequently if they played a series of one minute games....that's fair enough, and quite understandable, but the problem here is that this assumes that the mavs and celts would start each game tied 0-0 and alternate who gets first possession.

the problem is that games a very rarely tied with a jump ball down the stretch -- usually one team has an advantage...one team is jacking up three's while the other team is getting free throws on the other end of the court.

i suppose someone could argue that taking an advantage into the final minutes is just luck, but that would be pushing the argument to the absurd, IMHO. but I digress.

the devil is always in the details.....looking closer at the mavs 16 *close* games...

14-2

Who would argue that it's any great luck for a team to win a game when they've got a two possession lead in the final minute? The mavs carried a two possession lead into the last minute of 4 of their wins.

10-2

I would expect a team that has a 2-3 point lead and possession at the one minute mark will win far more often than not...that is, if you've got a one basket lead and you'll probably get two possessions to the other teams one, you ought to win far more often than not -- fair enough?

mavs were 3-0 when they had a bucket and possession, and 0-1 when Golden State had a bucket and a possession.

7-1

Stats be damned, we're to the point where I think that a closer examination of the play-by-play is the only thing that means anything.

Houston (w80-77) Juwan Howard and Luther Head combine for 0-3, JET hits two ft's;
Minn (w94-93) Ricky Davis goes 0-2, Dirk hits two ft's, Randy Foye bangs a meaningless trey at the buzzer;
Sac (w106-104) Brad Miller drains a three, Dirk hits a jumper and two ft's, Kings never get an opportunity to take a game tying shot in the final minute;
Tor (w97-96) 6 different toronto raptors chase after Dirk, freeing J-HO right under the basket...dallas get's wild ass lucky winning this one;
Ind (w115-113) terry hits a clutch jumper, j-ho blocks granger's shot, and then dirk hits 3 ft's down the stretch;
Utah (w108-105) Damp stuffs Deron Williams, Dirk knocks down a couple ft's;
LAL (L98-101) Sasha friggin' Vujacic bangs a three, JET gets jobbed at the rim;
Pho (w101-99) Dirk cans the jumper w/ a second left...

so, what can I say about the endings of the really, really close games...

1. Dallas gets a lot of stops down the stretch;
2. The Rockets suck;
3. Dirk is the league's mvp; and
4. the Mavs would be 36-1 over the last 37 if the ref makes that damned call on smush parker.

Usually Lurkin
03-07-2007, 12:00 PM
The "close game" argument can be looked at by itself, but in Hollinger's model it all comes out in the wash by the end.
Sometimes for Hollinger, losing a close game on the road will increase a teams numerical rating, and the home team that won the game will have a reduction in numerical rating.

Usually Lurkin
03-07-2007, 12:04 PM
according to stein, his power-rankings were the most frequently viewed page on the espn website. espn desired to get more of this, and hence they desired to have a power-ranking which would be updated daily...enter Hollinger.
Viewership at the view has skyrocketed since Rosie took over. I bet it's almost up to Jerry Springer levels. You can be intruiged by crap, and you can even love it, but who respects it?

Usually Lurkin
03-07-2007, 12:09 PM
So, what your example does is justify exactly what Hollinger is trying to do. The team that goes from 12 up to 4 up at the buzzer is, according to his model, a *much* better team than the one that goes from 4 up to 4 down. As a full-season model, anyway.
I thought the point for Hollinger was that an 8 point difference between -4 and +4 isn't as weighty as one between +4 and +12. Am I wrong about Holinger's formula? Or is it that everyone else is assuming there should be a +/- factor that Hollinger argues should not be included?

chumdawg
03-07-2007, 12:59 PM
I thought the point for Hollinger was that an 8 point difference between -4 and +4 isn't as weighty as one between +4 and +12. Am I wrong about Holinger's formula? Or is it that everyone else is assuming there should be a +/- factor that Hollinger argues should not be included?I believe that Hollinger's formula takes into account only a team's average margin of victory, with no respect to how close any individual games are. Just simply add 'em all up and divide by games played.

He discussed the issue of "close games" as people started analyzing his results. I think where Hollinger is coming from is this:

1) A team that manages to regularly win by a good margin is performing better than a team that manages to regularly win a close margin (with no respect to actual wins and losses).

2) A team that manages to regularly win by a close margin is performing better than a team that manages to regularly lose by a close margin.

The latter part IS included in Hollinger's methodology, and I think that has been lost on some of the people who have criticized him.

Imagine a game that is tied with a minute to play. Let's say that it can go one of two ways: win by four or lose by four. (Obviously it could go all kinds of other ways, but let's use these as "average" outcomes, since they are in keeping with the question posed upthread.) Hollinger seems to believe that it's basically down to chance whether the team wins or loses. This is the part that gets in a lot of people's craw, and I agree that it's far too simplistic an assumption.

But the rub is, that point is absolutely immaterial to what Hollinger is trying to accomplish in his formula. He is trying to measure how well a team is PLAYING, not how well they are winning. (If you want to know that, just look at the standings.) So, in this tie-game-with-a-minute-left scenario, if the team in question can manage to come out on the plus-four side the great majority of the time, that team WILL benefit a great deal in Hollinger's formula.

So, it's not that Hollinger dismisses winning close games. It's that he takes a more meta-view of a team's performance.

And the road/home adjustment is either A) not part of his formula, or B) washes out in the end, once a team plays 41 of each.

LRB
03-07-2007, 10:15 PM
Hollinger at best is nothing more than a snakeoil salesman. At worst he's a complete boob.

chumdawg
03-08-2007, 12:19 AM
Hollinger at best is nothing more than a snakeoil salesman. At worst he's a complete boob.Do you disagree with anything in his power rankings besides the Mavs not being an ultra-clear #1 above all other teams?

mkat
03-08-2007, 12:29 AM
the one question i have of hollinger is if these mavericks will be one of "those" teams...one of the greatest teams in the history of basketball. he said earlier in the year that this year's Suns team would be in one of his columns. I wonder if the mavs are lucky enough to be qualified that way by him now.

wmbwinn
03-08-2007, 01:27 AM
It is silly, and if ESPN/Hollinger were more interested in statistical models than website hits they'd acknowledge as much.

To repeat myself yet again and again...

models such as these do a lousy job of telling you whether the mavs will beat the suns in a 7 game series, but they do a great job of telling you that the Spurs would beat the Charlotte Bobcats in 7 game series. That is...

--they're great for telling you things which you already know, but don't care about...
--they're lousy for telling you things which you don't know, but do care about....



color me entirely unconvinced that winning close games is luck. That is, I am ignorant of any compelling argument or statistical evidence that this is in fact true. anyhoo....

chumdawg above makes a good illustration when he notes that the Celtics would beat the Dallas quite frequently if they played a series of one minute games....that's fair enough, and quite understandable, but the problem here is that this assumes that the mavs and celts would start each game tied 0-0 and alternate who gets first possession.

the problem is that games a very rarely tied with a jump ball down the stretch -- usually one team has an advantage...one team is jacking up three's while the other team is getting free throws on the other end of the court.

i suppose someone could argue that taking an advantage into the final minutes is just luck, but that would be pushing the argument to the absurd, IMHO. but I digress.

the devil is always in the details.....looking closer at the mavs 16 *close* games...

14-2

Who would argue that it's any great luck for a team to win a game when they've got a two possession lead in the final minute? The mavs carried a two possession lead into the last minute of 4 of their wins.

10-2

I would expect a team that has a 2-3 point lead and possession at the one minute mark will win far more often than not...that is, if you've got a one basket lead and you'll probably get two possessions to the other teams one, you ought to win far more often than not -- fair enough?

mavs were 3-0 when they had a bucket and possession, and 0-1 when Golden State had a bucket and a possession.

7-1

Stats be damned, we're to the point where I think that a closer examination of the play-by-play is the only thing that means anything.

Houston (w80-77) Juwan Howard and Luther Head combine for 0-3, JET hits two ft's;
Minn (w94-93) Ricky Davis goes 0-2, Dirk hits two ft's, Randy Foye bangs a meaningless trey at the buzzer;
Sac (w106-104) Brad Miller drains a three, Dirk hits a jumper and two ft's, Kings never get an opportunity to take a game tying shot in the final minute;
Tor (w97-96) 6 different toronto raptors chase after Dirk, freeing J-HO right under the basket...dallas get's wild ass lucky winning this one;
Ind (w115-113) terry hits a clutch jumper, j-ho blocks granger's shot, and then dirk hits 3 ft's down the stretch;
Utah (w108-105) Damp stuffs Deron Williams, Dirk knocks down a couple ft's;
LAL (L98-101) Sasha friggin' Vujacic bangs a three, JET gets jobbed at the rim;
Pho (w101-99) Dirk cans the jumper w/ a second left...

so, what can I say about the endings of the really, really close games...

1. Dallas gets a lot of stops down the stretch;
2. The Rockets suck;
3. Dirk is the league's mvp; and
4. the Mavs would be 36-1 over the last 37 if the ref makes that damned call on smush parker.


That is a great post. My main difficulty with Hollinger's model is that hypothesis that games decided by 5 or less points are luck or 50/50 deals. That is the statistical assumption or hypothesis (and Hollinger may have very good multi year support for his statement but it isn't true for these Mavs). The most repeated finding in your play by play at game end in close games was that the Dallas defense was awesome and the Mavs hit their free throws when fouled.
Half of the close games weren't reviewed due to the Mavs having enough of a lead with minimal time left to be boring for evaluation (but that indicates that in half of the close games, that luck or a 50/50 proposition is again a falsehood).
So, in conclusion the Mavs win their close games by:
1)Having enough of a lead to control the winding up scene
2)DEFENSE and free throws. In the 7 evaluated games, the opposition shot a combined 2 of 7 or 3 of 8 if you count the one loss to Sasha's insane shooting night. There were 2 additonal shots that were blocked which, if counted, would be 2 of 9 or 3 of 10. Dirk, Terry, and JHo hit shots and closed out games.

Now, we know why Hollinger's model doesn't fit the Mavs specifically as it relates to his assumption about the 50/50 split of close games...

One game was luck (Toronto game).

LRB
03-08-2007, 02:55 AM
Do you disagree with anything in his power rankings besides the Mavs not being an ultra-clear #1 above all other teams?

I think that his formula probably is right more times than it's not right. I don't have so much a problem with the Mavs not being ranked #1 as I do with Hollingers idiotic defenses of his formula as being right. The man doesn't want to publicly admit that his formula is imperfect, even though any person with an ounce of intelligence could tell you that it is NOT perfect.

Hey Maybe the playoffs will prove out that Dallas isn't the best team. I personally think that it is highly unlikely that this will happen unless major injuries occur to key Mav players. But even a diehard homer fan like myself will admit to the possibility that it could happen. Hollinger comes up with contradicting idiotic defense after idiotic defense of his "perfect" formular rather than admit the obvious that his formular is sometimes wrong.

Usually Lurkin
03-08-2007, 06:08 AM
Hey Maybe the playoffs will prove out that Dallas isn't the best team. ...
This is something that's starting to bother me about Hollinger's writeups. He keeps claiming that his margin of victory heavy formula is better for predicting future success, but I haven't seen his operational definition of "future success." Somewhere he said that his rankings would be the best way to bet on a future matchup between two individual teams. That seems definite enough, and testable. Elsewhere his defense was something along the lines of "so the high MOV team didn't win a championship that year, but the measurement is valid because they won it the next year." Which seems a little more vague to say the least.

alexamenos
03-08-2007, 10:14 AM
So, in conclusion the Mavs win their close games by:
1)Having enough of a lead to control the winding up scene
2)DEFENSE and free throws. In the 7 evaluated games, the opposition shot a combined 2 of 7 or 3 of 8 if you count the one loss to Sasha's insane shooting night. There were 2 additonal shots that were blocked which, if counted, would be 2 of 9 or 3 of 10. Dirk, Terry, and JHo hit shots and closed out games.

Now, we know why Hollinger's model doesn't fit the Mavs specifically as it relates to his assumption about the 50/50 split of close games...

One game was luck (Toronto game).

yeah...

defense and freethrows....and rebounding I might add....how remarkable is it that a team that plays good defense, makes its free throws and rebounds gets lucky quite often?

Usually Lurkin
03-18-2007, 07:41 PM
Spurs are still ahead of the Mavs. And look at Holl's numbers below. There is only one number on the whole list that's higher for the Spurs than the Mavs. And that's by a measly .9 points. Why does he even bother with a formula?
SPURS:
HOLLINGER RANK RATING RECORD MARGIN SOS REC (L10) MAR (L25%) SOS (L25%) STREAK DIV CONF
1 1111
109.027 46-20 +8.14 .504 8-2 +9.82 .475 Lost 2 2nd 3rd

MAVS:
HOLLINGER RANK RATING RECORD MARGIN SOS REC (L10) MAR (L25%) SOS (L25%) STREAK DIV CONF
2 2222
108.657 53-11 +8.05 .507 8-2 +11.69 .496 Won 2 1st 1st

I really wish I'd been saving his numbers 'cause I swear a few weeks ago the mavs were ahead on MOV, but were behind in the rankings. He claimed SOS was the difference then. And hasn't he also recently argued that recent MOV was very important in his mind?


Edit: if the rankings are still this way tomorrow, when today's win over Detroit is factored in, I'll have to find a crotch kicking gif to email him.

Dirkadirkastan
03-18-2007, 08:25 PM
Maybe Hollinger should integrate point differential over time. Then he'd be taking into account the differential at every second of the regular season. :D

Usually Lurkin
03-18-2007, 09:17 PM
man, the Rockets beat the Sixers by 50 points (124-71). Here's the Rockets entry at 8 (right behind the Pistons). How much might they move up with a single 50 point win?

HOLLINGER RANK RATING RECORD MARGIN SOS REC (L10) MAR (L25%) SOS (L25%) STREAK DIV CONF
8 8888
104.247 41-25 +4.83 .512 7-3 +3.06 .513 Won 3 3rd 5th
Prev Rnk: 8 | Prev Rat: 104.259 | Lst Game: @TOR, W 114-100 | Nxt 3: @PHI (Sun), IND (Tues), DET (Thur)

chumdawg
03-18-2007, 10:27 PM
Spurs are still ahead of the Mavs. And look at Holl's numbers below. There is only one number on the whole list that's higher for the Spurs than the Mavs. And that's by a measly .9 points. Why does he even bother with a formula?
SPURS:
HOLLINGER RANK RATING RECORD MARGIN SOS REC (L10) MAR (L25%) SOS (L25%) STREAK DIV CONF
1 1111
109.027 46-20 +8.14 .504 8-2 +9.82 .475 Lost 2 2nd 3rd

MAVS:
HOLLINGER RANK RATING RECORD MARGIN SOS REC (L10) MAR (L25%) SOS (L25%) STREAK DIV CONF
2 2222
108.657 53-11 +8.05 .507 8-2 +11.69 .496 Won 2 1st 1st

I really wish I'd been saving his numbers 'cause I swear a few weeks ago the mavs were ahead on MOV, but were behind in the rankings. He claimed SOS was the difference then. And hasn't he also recently argued that recent MOV was very important in his mind?


Edit: if the rankings are still this way tomorrow, when today's win over Detroit is factored in, I'll have to find a crotch kicking gif to email him.You know what, UL? I worked it out for myself over here--warning, it takes a while--and San Antonio's number doesn't match what I got. Dallas's does. I'm getting San Antonio at about 108.74. I will look into this further to see if I can find out where I'm wrong (IF I'm wrong). Odd that I get Dallas's number exactly, but not San Antonio's.

Some of the discrepancies are down to rounding errors, since Hollinger only goes to so many decimal places. I can get exact numbers on my own when it comes to margin of victory, but I can't do anything about strength of schedule. (It should be noted, though, that strength of schedule plays a very, very small role in this comparison.)

Something seems a little fishy here...

chumdawg
03-18-2007, 10:35 PM
I think it has to do with him using the last 25% of games played, as opposed to last ten games, later in the season. But I'm still not getting the Spurs at over 109.

Usually Lurkin
03-19-2007, 05:59 AM
wow. thanks chumdawg. that's something I've been wanting to do, but haven't had the time.

Here's the top 3 today (with yesterday's games factored in). San Antonio's number went up even though they didn't play (is sos calculated based on the current standings, rather than how teams were scored at the time of the game?) And Houston moved up a whopping 5 places based on the results of one game. Dallas's obscene win % for games under 5 points is chance, but Houston and the Sixers getting together for one blowout is so Truthful that it's worth a whole season of games for yesterday's 3-6 teams.

SAN ANTONIO
HOLLINGER RANK RATING RECORD MARGIN SOS REC (L10) MAR (L25%) SOS (L25%) STREAK DIV CONF
1 1111 109.033 46-20 +8.14 .504 8-2 +9.82 .476 Lost 2 2nd 3rd
Prev Rnk: 1 | Prev Rat: 109.027 | Lst Game: BOS, L 85-91 | Nxt 3: IND (Wed), DET (Fri), @SEA (Sun)


DALLAS
HOLLINGER RANK RATING RECORD MARGIN SOS REC (L10) MAR (L25%) SOS (L25%) STREAK DIV CONF
2 2222 108.753 54-11 +7.98 .510 8-2 +11.24 .506
Prev Rnk: 2 | Prev Rat: 108.657 | Lst Game: @DET, W 92-88 | Nxt 3: @NY (Tues), @CLE (Wed), @BOS (Fri)

HOUSTON
HOLLINGER RANK RATING RECORD MARGIN SOS REC (L10) MAR (L25%) SOS (L25%) STREAK DIV CONF
3 3333 105.513 42-25 +5.51 .511 7-3 +4.76 .515 Won 3 3rd 5th
Prev Rnk: 8 | Prev Rat: 104.247 | Lst Game: @PHI, W 124-74 | Nxt 3: IND (Tues), DET (Thur), @NOK (Sun)

edit: I just saw that SOS is calculated on w-l percentages. It's not clear, but it looks like it's current w-l. It should be sos at the time of games played. And I think it's hilarious that the stregth of a team is in it's w-l record, which everyone knows isn't as good a measure as MOV ;)

bobatundi
03-19-2007, 07:33 AM
man, the Rockets beat the Sixers by 50 points (124-71). Here's the Rockets entry at 8 (right behind the Pistons). How much might they move up with a single 50 point win?


I half-expected them to move up to #1; instead, they jump from 8th best team in the league to 3rd best on the strength of one win over one of the worst teams in the league. Which just further proves how stupid Hollinger's system is. You have to control for outliers when building a model, otherwise they can dominate your model and give you nonsensical predictions--like, Houston is the 3rd best team in the league right now.

alexamenos
03-19-2007, 09:55 AM
I half-expected them to move up to #1; instead, they jump from 8th best team in the league to 3rd best on the strength of one win over one of the worst teams in the league. Which just further proves how stupid Hollinger's system is. You have to control for outliers when building a model, otherwise they can dominate your model and give you nonsensical predictions--like, Houston is the 3rd best team in the league right now.

i vaguely recall that one week ago the number 4 ranked team beat the number 2 ranked team and kicked the ever-livin' shit out of the number 3 ranked team before getting creamed by the number 8 ranked team. Hollinger's model may not be remotely useful, but at least it generates chatter.

Dirkadirkastan
03-19-2007, 11:09 AM
I think we have enough evidence against Hollinger that we can write a dissertation now. Was that our goal from the beginning?

Flacolaco
03-19-2007, 11:34 AM
I think we have enough evidence against Hollinger that we can write a dissertation now. Was that our goal from the beginning?

Oh... I had something in mind with more pitchforks and torches and an angry mob....but that sounds good too.

Dirkadirkastan
03-19-2007, 12:51 PM
Oh... I had something in mind with more pitchforks and torches and an angry mob....but that sounds good too.

That sounds great, but do you really need this much evidence to arrange that? Seems like saying "Spurs are #1" would be enough to galvanize the mob, do you agree?

Maybe you're onto something. I could have been looking at this from the wrong angle all along...

LRB
03-19-2007, 12:56 PM
I think at best Hollingers fomula should be + or - 3 to 5 places. But jumping Houston up so far on beating up a lottery bound team is just stupid. What still kills me about holinger's formula is that if you're up 30 on a team going into the 4th, you would have a better team if you leave your starters in and try for a 40 or 50 MOV than if you rest the starters and let the scrubs get some burn and maybe only win by 20 or 25. The 1st is poor sportsmanship and just plain stupid coaching. The 2nd is good sportsman ship and very smart coaching. But Hollinger's formula rewards the stupid coach's and penalizes the smart coach's.

chumdawg
03-19-2007, 06:39 PM
wow. thanks chumdawg. that's something I've been wanting to do, but haven't had the time.

Here's the top 3 today (with yesterday's games factored in). San Antonio's number went up even though they didn't play (is sos calculated based on the current standings, rather than how teams were scored at the time of the game?) And Houston moved up a whopping 5 places based on the results of one game. Dallas's obscene win % for games under 5 points is chance, but Houston and the Sixers getting together for one blowout is so Truthful that it's worth a whole season of games for yesterday's 3-6 teams.

SAN ANTONIO
HOLLINGER RANK RATING RECORD MARGIN SOS REC (L10) MAR (L25%) SOS (L25%) STREAK DIV CONF
1 1111 109.033 46-20 +8.14 .504 8-2 +9.82 .476 Lost 2 2nd 3rd
Prev Rnk: 1 | Prev Rat: 109.027 | Lst Game: BOS, L 85-91 | Nxt 3: IND (Wed), DET (Fri), @SEA (Sun)


DALLAS
HOLLINGER RANK RATING RECORD MARGIN SOS REC (L10) MAR (L25%) SOS (L25%) STREAK DIV CONF
2 2222 108.753 54-11 +7.98 .510 8-2 +11.24 .506
Prev Rnk: 2 | Prev Rat: 108.657 | Lst Game: @DET, W 92-88 | Nxt 3: @NY (Tues), @CLE (Wed), @BOS (Fri)

HOUSTON
HOLLINGER RANK RATING RECORD MARGIN SOS REC (L10) MAR (L25%) SOS (L25%) STREAK DIV CONF
3 3333 105.513 42-25 +5.51 .511 7-3 +4.76 .515 Won 3 3rd 5th
Prev Rnk: 8 | Prev Rat: 104.247 | Lst Game: @PHI, W 124-74 | Nxt 3: IND (Tues), DET (Thur), @NOK (Sun)

edit: I just saw that SOS is calculated on w-l percentages. It's not clear, but it looks like it's current w-l. It should be sos at the time of games played. And I think it's hilarious that the stregth of a team is in it's w-l record, which everyone knows isn't as good a measure as MOV ;)I've been at this for about an hour, trying to figure out how in the hell he is getting his final numbers. I get:

DAL 108.9170
SAS 108.9019

I'll keep looking at it, but I can't understand what is going on. I'm using the stats from nba.com to get precise MARG numbers, so that I don't have to worry about any rounding errors. But even using his rounded numbers, I still don't get his results. And even if I adjust his SOS numbers downward by as much as I can and stay within the range of his rounded numbers, I can't get Dallas as low as he has them or San Antonio as high as he has them.

I've even tried different varieties of his .67 and .33 coefficients (like .666667 or, my preference, 2/3) and that doesn't help either.

I'm starting to think this is a sham.

jthig32
03-19-2007, 07:26 PM
I've been at this for about an hour, trying to figure out how in the hell he is getting his final numbers. I get:

DAL 108.9170
SAS 108.9019

I'll keep looking at it, but I can't understand what is going on. I'm using the stats from nba.com to get precise MARG numbers, so that I don't have to worry about any rounding errors. But even using his rounded numbers, I still don't get his results. And even if I adjust his SOS numbers downward by as much as I can and stay within the range of his rounded numbers, I can't get Dallas as low as he has them or San Antonio as high as he has them.

I've even tried different varieties of his .67 and .33 coefficients (like .666667 or, my preference, 2/3) and that doesn't help either.

I'm starting to think this is a sham.

He had this to say in his chat today:

Chris (Cibolo, TX): Please explain, how Dallas leads in all but one category of your rankings and still falls behind the Spurs?

John Hollinger: 1) They're trailing in the most important category, season point differential, and 2) there's a category you don't see, home games vs. road games, that also works in San Antonio's favor.

dude1394
03-19-2007, 08:10 PM
Pretty amazing...Houston blows out someone by what 50 and they move to 3rd. Really kinda wacky.

But he's so right on about the mvp.

chumdawg
03-19-2007, 09:22 PM
Thiggy: I'm using his formula exactly as he published it, and just plugging in the data. The formula does account for home/away games (else Dallas would be well out in front). What he meant in his response to that poster, I believe, is that the numbers ESPN publishes along with his daily rankings do not include the data about home/away games.

Either his formula has more to it than what they have published, or they are tweaking the numbers.

Flacolaco
03-19-2007, 09:40 PM
Thiggy: I'm using his formula exactly as he published it, and just plugging in the data. The formula does account for home/away games (else Dallas would be well out in front). What he meant in his response to that poster, I believe, is that the numbers ESPN publishes along with his daily rankings do not include the data about home/away games.

Either his formula has more to it than what they have published, or they are tweaking the numbers.

I wonder if enough people e-mail him about it, he'll eventually put out some sort of "proof" to show the math? Because apparently no one can figure it out (with the exception of Spurs fans)

jthig32
03-19-2007, 09:52 PM
Thiggy: I'm using his formula exactly as he published it, and just plugging in the data. The formula does account for home/away games (else Dallas would be well out in front). What he meant in his response to that poster, I believe, is that the numbers ESPN publishes along with his daily rankings do not include the data about home/away games.

Either his formula has more to it than what they have published, or they are tweaking the numbers.

Yeah, I realize that. I wasn't posting that as a "here's why they're different". Was just wanting you make sure you saw that at least some other people are questioning his results.

chumdawg
03-19-2007, 10:07 PM
Yeah, I realize that. I wasn't posting that as a "here's why they're different". Was just wanting you make sure you saw that at least some other people are questioning his results.Right. I gotcha. However, the way that question was posed to him, his was a legitimate answer.

Something is off, though. If I find enough inclination, I may check the data on all the teams and see if I can figure out why there are discrepancies.

mary
03-19-2007, 11:31 PM
If anyone else wants play with Hollingers numbers, just set up a spreadsheet as shown below and copy and paste the formula into column K. The last number in the formula is 25% of total games played, rounded up. You could use (H2+I2) for the denominator, - I just manually entered the number.

Edit, here is my spreadsheet, and here is the formula -

Chum, do you mind writing out your formula, so I can use it? (Edited: nevermind, once I added a home game to Dallas, a road game to SA and changed the last denominator to 17, I pretty much got your numbers...very close anyway.)

As you can see, these don't match up with Hollinger's either, so it'll be interesting to find out what is going on.

http://home.comcast.net/%7Egolden75/Image1.gif

+(((B2-0.5)/0.037)*0.67) + (((D2-0.5)/0.037)*0.33) + 100 + (0.67*(C2+(((G2-F2)*3.5)/(J2))) +(0.33*(E2+(((I2-H2)*3.5)/(17)))))

alexamenos
03-20-2007, 09:52 AM
mary -- Have you checked the MOVs and SOSs as reported on the ESPN/Hollinger Web-page? I'd guess there is a fair possibility of a disparity between those displayed and the correct inputs (used by Hollinger)....I seem to recall encountering a problem like this when I replicated his rankings, anyway.

chumdawg
03-20-2007, 09:54 AM
That's sweet, Mary. I don't know how to paste mine in like you did. Cool.

I actually didn't enter the entire formula in one cell, primarily because I was interested in looking at the different "components" of the rating, if you will. For example, I found that the SOS part of it was miniscule.

Other than that, though, we did it the exact same way, it looks like. I went and got the actual numbers for the MARGins, like you did, rather than using the rounded forms.

I think the only difference is that I used the fractions (1/3 and 2/3) rather than .33 and .67.

I wish I had the inclination to check the rest of the teams, or at least a few of them. When I get a slow day at work, maybe. I'm pretty sure I recall getting his exact same number for the Mavs a couple days ago, and noticing the San Antonio number was off. If I did get his exact same number at that time, then you would think he's using the same formula we are (the one he published).

chumdawg
03-20-2007, 09:55 AM
mary -- Have you checked the MOVs and SOSs as reported on the ESPN/Hollinger Web-page? I'd guess there is a fair possibility of a disparity between those displayed and the correct inputs (used by Hollinger)....I seem to recall encountering a problem like this when I replicated his rankings, anyway.The MOV's are easy to calculate on your own, and the SOS's have a tiny contribution to the end number. I think he's fudging it for the sake of argument and interest.

mary
03-20-2007, 10:01 AM
mary -- Have you checked the MOVs and SOSs as reported on the ESPN/Hollinger Web-page? I'd guess there is a fair possibility of a disparity between those displayed and the correct inputs (used by Hollinger)....I seem to recall encountering a problem like this when I replicated his rankings, anyway.
Yes, both Chum and I calculated our own MOV.

I didn't do SOS, as that would take a REALLY long time.

However I did use the Goal Seek function to see if I could arrive at Hollinger's numbers by tweaking the SOS. But that was impossible to do without changing the actual value of SOS.

Kudos to Chum for uncovering this. I think you should send Hollinger an e-mail.

chumdawg
03-20-2007, 10:37 AM
Mary, what did Goal Seek tell you the values of SOS would have to be?

I don't have much time this morning to look at it again, but I recall that it looked like you would have to change the SOS quite a bit to get to his numbers. And of course, you go in opposite directions with the two teams (Mavs down, Spurs up).

I guess it's conceivable that they published the wrong SOS numbers. Seems doubtful, though.

alexamenos
03-20-2007, 10:38 AM
Yes, both Chum and I calculated our own MOV.

I didn't do SOS, as that would take a REALLY long time.

i know it would take a long time, but I nonetheless suspect that this is where the difference is...

....checking sos of dallas' last 25% -- it should be 0.494 (17 games) rather than .509 (last 16 games)...that change alone will get you under 108.8 for the Mavs.

which takes me back to the point I was trying to make --> here (http://www.dallas-mavs.com/vb/showthread.php?p=689407&highlight=deconstructing#post689407).

the relative Hollinger rankings of Dallas and San Antonio may well hinge upon the exceedingly trivial and arbitrary question of whether we should include Memphis's .294 season record in the "recent" part of the equation....

.......this consideration is considerably less important than whether San Antonio can rebound adequately with the small line-up of Parker-Ginobili-Finley-Bowen-Duncan lineup on the court.

cheers

alexamenos
03-20-2007, 10:45 AM
I guess it's conceivable that they published the wrong SOS numbers. Seems doubtful, though.

i'm reasonably certain this is the case -- dallas' sos25 is published as .509, which is true for the last 16 games but not the last 17 while the mar25 is 11.24, which is true for the last 17 but not the last 16.

i quite reasonably suspect that the stats published on espn and the data Hollinger uses are not linked in anyway, and such discrepancies are common throughout the rankings....

i'm sure espn is confident that only persons who happen to be stats geeks and mildly obsessive fans of the wrongly ranked number 2 team might ever uncover such a thing....and hence they're not terribly concerned about the disparity.

mary
03-20-2007, 10:53 AM
Mary, what did Goal Seek tell you the values of SOS would have to be?

I don't have much time this morning to look at it again, but I recall that it looked like you would have to change the SOS quite a bit to get to his numbers. And of course, you go in opposite directions with the two teams (Mavs down, Spurs up).

I guess it's conceivable that they published the wrong SOS numbers. Seems doubtful, though.
First I changed by .33 and .67 to 1/3 and 2/3, so our numbers match up now.

To match Hollingers numbers, I changed the value of overall SOS. It gave me this:

Mavs .500896
Spurs .511276

So it would change the values of the SOS. Theoretically, I guess the difference could be made up in the SOS for the L25.

mary
03-20-2007, 10:57 AM
i'm reasonably certain this is the case -- dallas' sos25 is published as .509, which is true for the last 16 games but not the last 17 while the mar25 is 11.24, which is true for the last 17 but not the last 16.

i quite reasonably suspect that the stats published on espn and the data Hollinger uses are not linked in anyway, and such discrepancies are common throughout the rankings....

i'm sure espn is confident that only persons who happen to be stats geeks and mildly obsessive fans of the wrongly ranked number 2 team might ever uncover such a thing....and hence they're not terribly concerned about the disparity.

It also seems weird to me that Hollnger would allow incorrect info to be published on his rankings.

At the same time, its seems logical that the only discrepency can be in the SOS calculation. Everything else is pretty black and white.

chumdawg
03-20-2007, 10:57 AM
Ah, I see. I'm sure that's the deal, then. They just didn't update the SOS numbers. Well, that's pretty ridiculous, if you ask me.

If you plug in .5069 as the SOS for Dallas, and your .494 for the last 25%, you get Hollinger's number. I guess that's not as big a stretch as I thought at first.

chumdawg
03-20-2007, 11:00 AM
It also seems weird to me that Hollnger would allow incorrect info to be published on his rankings.

At the same time, its seems logical that the only discrepency can be in the SOS calculation. Everything else is pretty black and white.Yeah, same here. Ah, well. They still should be more careful about publishing the right numbers. Else someone might be tempted to conclude that Hollinger is a shill!

mary
03-20-2007, 11:01 AM
Yeah, same here. Ah, well. They still should be more careful about publishing the right numbers. Else someone might be tempted to conclude that Hollinger is a shill!

It was way more exciting to think Hollinger is a shill.

As it turns out, he probably just has a lazy intern.

:D

alexamenos
03-20-2007, 11:08 AM
It was way more exciting to think Hollinger is a shill.

As it turns out, he probably just has a lazy intern.

:D

bear in mind that the purpose of the Hollinger Daily Rankings is to drive daily traffic to espn.com, not to satisfy the anal and obsessive tastes of wee math geeks....

mffl03
03-21-2007, 09:16 AM
3/21/07

So Hollinger just listed us as his #1, I bet if we lose tonight we go down to number 2 again...

then again who really gives a flying ::bleep:: what this guy says

Underdog
03-21-2007, 09:18 AM
3/21/07

So Hollinger just listed us as his #1, I bet if we lose tonight we go down to number 2 again...

then again who really gives a flying ::bleep:: what this guy says


What are you talking about??? We can win tonight and STILL be #2 tomorrow - Hollinger's system is based on whatever results he wants to achieve... :rolleyes:

dirno2000
03-21-2007, 09:20 AM
then again who really gives a flying ::bleep:: what this guy says

we do?

mary
03-21-2007, 09:27 AM
3/21/07

So Hollinger just listed us as his #1, I bet if we lose tonight we go down to number 2 again...

then again who really gives a flying ::bleep:: what this guy says

There are at least TWO Hollinger threads in the NBA section.

I've had enough whining about this.

Its a "bleeping" computer ranking.

Its based on a formula.

Its not subjective.

I think the analysis of his formula makes a pretty good discussion, and its certainly subject to valid criticism - but Hollinger doesn't have an anti-Mavs agenda.

This is the most over-complained about power ranking in the history of this forum.

shaw-xx
03-21-2007, 09:46 AM
I've officially ignored Hollinger's power rankings. :D

alexamenos
03-21-2007, 10:49 AM
i like espn's link to hollinger's ratings has for weeks said something like "Dallas and San Antonio continue to battle it out for the top spot"....

....as if the mavs are going into the game thiking, 'yeah, if we can beat the knicks tonight by 13 or more we'll get the number one spot in a computer poll of dubious value.'

kriD
03-21-2007, 10:54 AM
There are at least TWO Hollinger threads in the NBA section.

I've had enough whining about this.

Its a "bleeping" computer ranking.

Its based on a formula.

Its not subjective.

I think the analysis of his formula makes a pretty good discussion, and its certainly subject to valid criticism - but Hollinger doesn't have an anti-Mavs agenda.

This is the most over-complained about power ranking in the history of this forum.
Best post about this topic.

"You must spread some Reputation around before giving it to mary again."

LRB
03-21-2007, 11:41 AM
There are at least TWO Hollinger threads in the NBA section.

I've had enough whining about this.

Its a "bleeping" computer ranking.

Its based on a formula.

Its not subjective.

I think the analysis of his formula makes a pretty good discussion, and its certainly subject to valid criticism - but Hollinger doesn't have an anti-Mavs agenda.

This is the most over-complained about power ranking in the history of this forum.

I have tons of issues with the formula itself. But you're completely right in that the formula is objective. In no way is Hollinger biased towards the Mavs. Any bias he might have is towards self promotion of his formula, nothing more and nothing less.

Usually Lurkin
03-21-2007, 12:17 PM
This is the most over-complained about power ranking in the history of this forum.
For some of us (judging by the amount of time some of ya'll have put in, I'm not the only one) it's also the coolest.


In no way is Hollinger biased towards the Mavs. Any bias he might have is towards self promotion of his formula, nothing more and nothing less.
Yeah, my only beef with Hollinger himself (and it's pretty big) is that he's not discussing Stats fairly, not that he's treating the Mavs unfairly.

Dirkadirkastan
03-21-2007, 12:32 PM
So what if it's a formula. Some of us are questioning whether the posted numbers are even correct.

mary
03-21-2007, 12:41 PM
So what if it's a formula. Some of us are questioning whether the posted numbers are even correct.

The numbers themselves are correct. The only problem was that the SOS being published on the rankings was not being updated - as discovered by alexiconamos (sp) and chumdawg.

alexamenos
03-21-2007, 01:10 PM
...by alexiconamos (sp)

nyuk, nyuk...

so often mispelled, but this is one of the better ones....a-lexicon-amos.;)

(i answer to alex, btw...."alexamenos (http://altreligion.about.com/library/glossary/symbols/bldefsalexamenos.htm)" is a tragically-comical figure in a piece of ancient anti-Christian graffiti. alexamenos is a fool, but possibly an earnest fool...he looks kind of like a first century bart simpson)

MartinGT
03-22-2007, 03:56 AM
Here are the last 15 NBA champs, along with their league wide ranking in win percentage and point differential:

Year Champ Win % Point Differential
2006 Heat 5th 5th
2005 Spurs 2nd 1st
2004 Pistons 6th 2nd
2003 Spurs 1st 3rd
2002 Lakers 2nd 2nd
2001 Lakers 2nd 7th
2000 Lakers 1st 1st
1999 Spurs 1st 1st
1998 Bulls 1st 1st
1997 Bulls 1st 1st
1996 Bulls 1st 1st
1995 Rockets 10th 10th
1994 Rockets 2nd 6th
1993 Bulls 3rd 2nd
1992 Bulls 1st 1st

Average rank: win %: 2.6, point diff.: 2.9

Over that time, winning percentage has an edge over point differential in predicting the champion, but its very slight.

http://www.dallas-mavs.com/vb/showthread.php?t=28446&page=2&highlight=probability

post #54

Another nice history exam that shows that winning percentage is a better predictor thatn point differential...

that refutes Hollinger squarely....

Where are the Mavs on that list??

LOL!!!!!!

Thespiralgoeson
03-22-2007, 05:45 AM
Where are the Mavs on that list??

LOL!!!!!!

wtf??

What point exactly are you trying to make here? That the Mavs haven't won a championship? If so... how exactly is that relevant to a discussion about John Hollinger's ranking system? Jesus... even for a troll, you're dense.

mffl03
03-22-2007, 09:00 AM
HEEEEEEEEEEEEEEY. Mavs are back in second place after winning 3 in a row, especially after beating the top 2 teams in the East.....

someone called it here....pat yourself on the back

LRB
03-22-2007, 11:56 AM
wtf??

What point exactly are you trying to make here? That the Mavs haven't won a championship? If so... how exactly is that relevant to a discussion about John Hollinger's ranking system? Jesus... even for a troll, you're dense.


Just goes to show you that you can't take the dumb out of the ass.

Mavinator
03-22-2007, 02:41 PM
The thing about statistics is its basically a "gray area" science- the mathematics of it are concrete, but both the assumptions you work with to construct your model and the conclusions you draw from it are open to discretion and interpretation. That is to say, stats are great for predicting dice rolls and PCB concentrations in fish populations, but when you're dealing with something as incredibly complex as basketball, with hundreds of thousands of constantly changing variables involved in every single game, and with the relationships between these variables unknown except to the subjective predictions of sportswriters, any prediction drawn from statwork is basically bupkus. You could prove anything if you put your mind to it and picked the right numbers.

That being said, from what I can tell Hollinger's "predictive model" is more a case of applied eighth-grade algebra than actual probability work

u2sarajevo
03-22-2007, 02:51 PM
What's up Mavinator.... tell your brother he needs to visit.

On the Hollinger ratings I saw something that I keep seeing from fans:
With 14 games left for the Rockets, I'm hoping they go 10-4 and pass up the Jazz to secure homecourt. They have to start tonight against the Pistons!But the Rockets to pass up the Jazz for homecourt in 1st round would have to get past the Spurs right? Or is the rule apply to top 4 records regardless of divisional affiliation? Is it possible to have 1.Dallas 2. Spurs 3. Phoenix 4. Houston 5. Utah?

Someone?

Usually Lurkin
03-22-2007, 02:55 PM
any prediction drawn from statwork is basically bupkus. You could prove anything if you put your mind to it and picked the right numbers. That's not true. it's a matter of specification. His formula has Spurs, Mavs, Suns as the top 3 and Memphis, Charlotte, and Atlanta in the bottom. That's at least close for an objective method of ranking teams from best to worst. If you are trying to rank teams from best to worst, you won't get something that puts Memphis, Charlotte, and Atlanta in the top 3 and Dallas, SanAntonio, and Phoenix in the bottom.

That being said, from what I can tell Hollinger's "predictive model" is more a case of applied eighth-grade algebra than actual probability work
The statistics would've been done before hand, and the formula the result of the statistics. As part of his "method" page, I'd like to see how he came up with the formula, not just what the formula looks like.

Flacolaco
03-22-2007, 02:56 PM
What's up Mavinator.... tell your brother he needs to visit.

On the Hollinger ratings I saw something that I keep seeing from fans:
But the Rockets to pass up the Jazz for homecourt in 1st round would have to get past the Spurs right? Or is the rule apply to top 4 records regardless of divisional affiliation? Is it possible to have 1.Dallas 2. Spurs 3. Phoenix 4. Houston 5. Utah?

Someone?

It was my understanding that the division leaders were guaranteed home court (one of the top 4 spots)

alexamenos
03-22-2007, 03:03 PM
....His formula has Spurs, Mavs, Suns as the top 3 and Memphis, Charlotte, and Atlanta in the bottom....

which, in my view, further illustrates the (lack of) usefulness of such formulas. they're great for telling you things which are unbelievably obvious, like mavs, spurs, suns>grizz, b'cats, hawks. they're useless for telling you things which you'd really like to know, like whether mavs>spurs>suns.

LRB
03-22-2007, 03:28 PM
which, in my view, further illustrates the (lack of) usefulness of such formulas. they're great for telling you things which are unbelievably obvious, like mavs, spurs, suns>grizz, b'cats, hawks. they're useless for telling you things which you'd really like to know, like whether mavs>spurs>suns.


Hollinger's formula was pretty worthless in predicting that Miami would go on the role it did after Dshoulder went down with injury. In truthfulness I don't see how it's very predictive of what will happen in the future at all. I don't that any pure statistical formula can be highly accurate. That are too many variables in the NBA game, and many of those variables just don't have enough statisitical data to analyse. Hollinger's self promotion of the formula as being near 100% accurate is just flat out self serving BS. He rates the Sonics higher than the Lakers, Clippers, Warriors, Kings, and Hornets. Only the Sonics are hardly legit contenders for a playoff spot in the west while the other 5 are. How come Hollingers rating system failed so bad?

alexamenos
03-22-2007, 03:33 PM
Hollinger's formula was pretty worthless in predicting that Miami would go on the role it did after Dshoulder went down with injury. In truthfulness I don't see how it's very predictive of what will happen in the future at all....

as they saying goes....

predictions are tough, especially when they're about the future. ;)

LRB
03-22-2007, 03:37 PM
as they saying goes....

predictions are tough, especially when they're about the future. ;)

Looks like Hollinger's formula is worthless in showing which team won more in the past, likewise it's piss poor at showing which team will win in the future. But theorecically it's perfect at showing which team would win at the exact moment in time which it's published. Trouble is by the time my browser refreshes that momemnt in time has passsed by.

Mavinator
03-22-2007, 03:39 PM
That's not true. it's a matter of specification. His formula has Spurs, Mavs, Suns as the top 3 and Memphis, Charlotte, and Atlanta in the bottom. That's at least close for an objective method of ranking teams from best to worst. If you are trying to rank teams from best to worst, you won't get something that puts Memphis, Charlotte, and Atlanta in the top 3 and Dallas, SanAntonio, and Phoenix in the bottom.

You're saying his calculations are fairly accurate because Dallas, SA, and PHX are the best teams and Memphis etc are the worst teams, but that's an arbitrary claim in and of itself- a matter of opinion. What makes Dallas any better than Charlotte? Is it because we have more wins? And in that case, wouldn't win-loss record be a better predictive indicator of win-loss record than the Hollinger model?

What if instead of ranking "best team" by win-loss record, we defined the "best" team to be the team that misses the most shots? Then we could construct a model based on field goal percentage.

1. Indiana .435
2. Charlotte .440
3. Atlanta .441

And lo and behold, Charlotte has missed a fantastic 3121 shots this year, with Indiana and Atlanta not far behind with 3054 and 2998 misses respectively. Our predictive model is fairly accurate- a victory for statistics! But what does it really tell us?

Our model was designed to tell us the best team (i.e. the team that missed the most shots in the regular season) based on field goal percentage, and it was moderately accurate but it was obviously an incredibly stupid model. Hollinger's makes more sense to the casual reader, but the models are essentially the same. They are both drawn from arbitrarily chosen statistics, and they both draw conclusions based on an entirely subjective definition of "best team." In other words, the math is accurate, but the math is pointless- in the end the best team is who you say it is, right?


The statistics would've been done before hand, and the formula the result of the statistics. As part of his "method" page, I'd like to see how he came up with the formula, not just what the formula looks like.

Presumably this is the relevant bit of information that could lend some weight to his work, but its something he's never bothered to tell us. Did he develop his formula, test it against every team in the history of the NBA, and find that there was some statistically significant correlation between being number one in the Hollinger stat and eventually winning the title that year? And did he test a variety of other formulas against the whole of NBA history and determine that they were less effective estimators? Somehow, I highly doubt it.



Edit: Misspelled weight

alexamenos
03-22-2007, 03:48 PM
technically, I suppose that Hollinger would argue that his model is not supposed to predict head-head matchups, but rather which teams would win the most games if they played identical 82 game schedules starting tomorrow.

to me, that is rather useless information as long as playoff seedings are based on actual historic win-loss records and not predicted future win-loss records.

LonghornDub
03-22-2007, 05:04 PM
What's up Mavinator.... tell your brother he needs to visit.

On the Hollinger ratings I saw something that I keep seeing from fans:
But the Rockets to pass up the Jazz for homecourt in 1st round would have to get past the Spurs right? Or is the rule apply to top 4 records regardless of divisional affiliation? Is it possible to have 1.Dallas 2. Spurs 3. Phoenix 4. Houston 5. Utah?

Someone?

Hey u2,

You're right. Houston can't get into the top 4, because the 3 division leaders + the 4th team with the best record are guaranteed the top 4 spots. Unless the Rockets passed the Spurs AND the Jazz in the standings, such that they were the 2nd place team behind the Mavs in the division, then they can't be any higher than 5.

Usually Lurkin
03-22-2007, 07:42 PM
technically, I suppose that Hollinger would argue that his model is not supposed to predict head-head matchups, but rather which teams would win the most games if they played identical 82 game schedules starting tomorrow.
technically or not, he did literally claim that. In a defense of his ranking, he claimed that it would be better to bet on SA in a matchup between SA and Dallas. He's also argued that MOV is valid because sometimes the team with the highest MOV wins the championship the next year. So his argument is general, but he still hasn't let us in on what data he used to build his model.

What if instead of ranking "best team" by win-loss record, we defined the "best" team to be the team that misses the most shots? Then we could construct a model based on field goal percentage. I'd say you're not making an honest attempt at ranking teams on goodness, but that you are deliberately ranking teams on badness while changing the meanings of words. That's not statistics. That's semantics. If he's changing the meaning of Margin of Victory, or of Strength of Schedule, or changing his formula in order for some teams to be ranked higher than others, then he'd be doing what you did. Of course, if he were playing semantics in order to come out with a ranking that everyone thought was wrong, well he would be an idiot. However, if he defined those things beforehand according to some method (that we'd all like to see), and has stuck to it, then yeah, his ranking is doing "pretty good." Whether or not it is sufficiently good is up to any individual reader to decide. But because it's an objective, statistical model, it can be tweaked in an honest effort to rank teams by goodness, and do an unarguably better job next time.


And did he test a variety of other formulas against the whole of NBA history and determine that they were less effective estimators? Somehow, I highly doubt it.
I doubt his numbers are truly arbitrary, that he just pulled his model out of thin air, or that he didn't test anything. It is very easy to set an objective measure of goodness like win-loss, or playoff seeding, or playoff success, or head to head matchup, or whatever, and then to build a formula that will best fit past data. It's quite a bit harder to build one that will successfully predict future data. But because it's so easy to fit a model to past data, I would bet that he's done that. He just hasn't let us know what that past data is. My guess is he's being protective. If someone knows what data he's working on, then it makes it easier for them to do better than he did.

Maybe our stats friends here on the board who are fortunate enough to have time to play with it can search for a data set that his model fits.

jthig32
03-22-2007, 08:53 PM
Hey u2,

You're right. Houston can't get into the top 4, because the 3 division leaders + the 4th team with the best record are guaranteed the top 4 spots. Unless the Rockets passed the Spurs AND the Jazz in the standings, such that they were the 2nd place team behind the Mavs in the division, then they can't be any higher than 5.

But, they CAN get homecourt advantage over the Jazz. Homecourt is based on -record, not playoff seed.

chumdawg
03-22-2007, 10:01 PM
I'll try to chip in bit here on the discussion about Hollinger's ranking system. I'm no stats guru, but I do enjoy reading about what the stats gurus are doing when it comes to basketball. (From what I understand, they are in a way following in the footsteps of baseball stats work that is generally very well respected, even if the game of basketball is not so easy to quantify. One thing is for sure: guys are making money doing it, and teams--such as our Mavericks--are taking it seriously. So I do think it's worth following.)

The only thing that is at all questionable, the way I understand these issues, is Hollinger's methodology in devising his formula. In other words, his formula is an attempt to interpret statistical evidence and generate some conclusions from it. As Mavinator has described, that's usually where the rub is. But the data he's using is generally regarded to be bulletproof.

That data falls into two categories: A) Margin of Victory, and B) Home Court Advantage.

Take B first, because I think it's the most straightforward. Simply put, basketball stats guy have built databases comprising decades and decades of basketball games, and they have decided that the homecourt is worth 3.5 (give or take) points per game. I don't know all the ins and outs of how they arrived at that, but I know that they did it by simply observing a shitload of data. It's just an average, and like anything else it will have some variance, but evidently it holds up very well over the long run.

Now, the Margin of Victory measure is a little different, because it's not meant to be so empirical. This is the one that is stats-driven. But at its essence it is an indispensable tool, in many areas of basketball stats work. For example, it lies at the heart of even individual player performance measures like PER and Win Shares. It is, essentially, a measure of something like "success rate" at a micro level. It is used in modeling purposes to try and come up with a similar "success rate" on a macro level.

For example. Take a hypothetical league that consists of only two players, who play 82 games of one-on-one. One of those players is, say, Michael Jordan in his prime and the other is some ten-year-old kid. So these two play out the league, and according to the stats, for every two possessions back and forth, Jordan scores 1.9 points and the kid scores .1 points. (Occasionally the kid gets a lucky rebound, and occasionally he makes a circus shot.) The stats guys will take those numbers and try to predict a win-loss record for the two players.

Obviously, it's going to be 82-0. That's easy. But it's never so easy in the NBA.

Now imagine two teams, call them Dallas and San Antonio, playing 82 games against each other. Let's say the Mavs net .001 points for every exchange of possessions over the course of the season, and the Spurs net -.001. How would the model predict the win-loss record?

Probably not very well. The best the model could do would be 41-41 or maybe 42-40, I'm guessing. The real-world result could vary fairly widely (though one suspects that neither team would end up more than, say, ten games ahead).

So, would that be a failure of the model, which so accurately predicted Jordan's 82-0 over the kid--and probably fairly accurately predicted outcomes along the way as our two hypothetical participants got closer in ability?

Of course not. The model itself works just fine. But the model can't tell you things with much confidence as the data come so close together.

But the model knows very well that a team who, on average, nets a quarter-point per exchange of possessions will in fact kick the shit out of a team who on average gives up a quarter-point per exchange of possessions--in the long run, that is.

That's how they arrived at the Pythagoarean Win Expectation formula. They figured out that points-scored versus points-allowed, the micro element of wins and losses, do hold up pretty well on the macro level, over the long run. They have tweaked the value of the exponent as they have acquired more and more data, in an effort to make the model most closely match the body of evidence they have. I think currently they are using either 14.5 or 16 (I can't remember which), in a formula like this:

Points Scored^16, divided by

Points Scored^16 plus Points Allowed^16,

equals Expected Win percentage.

Some still use raw points, and some use Points per Possession in an attempt at a more accurate formula.

And that's really all Hollinger is using. He uses Home Court Advantage for this reason: Assume two identical teams, absolute mirror images of each other. Each team will win its home games by an average of 3.5 points (though not necessarily going 41-0) and lose its road games by an equal amount. Again, that's just something that's been observed over time, without any statistical model to stand on.

So Hollinger takes Margin of Victory, and tweaks it by 3.5 for each game according to where it was played.

With Strength of Schedule, he's doing the same thing. That's what the .037 coefficient is for. It's a conversion factor to go from win percentage to Margin of Victory.

The one thing Hollinger is doing that is at all debatable, if you accept his use of Margin of Victory as a valid statistical measure, is his weighting of recent games played. If he didn't do that, you would have nothing but Margin of Victory (though run through some filters) and we could argue about that. And again, that model isn't going to tell you a whole hell of a lot when you get two very, very closely-matched teams as far as this measure goes. Like the Mavs and Spurs.

But what it *might* tell you is that the Mavs and Spurs are a notch above the Suns and a good bit above everyone else.

That's why my buddy flying to Reno tonight will be shopping the futures odds for both the Mavs and Spurs, and playing a nice-sized bet on my behalf (on both teams, but in differing amounts). :)

Dirkadirkastan
03-22-2007, 11:04 PM
I'm pretty sure home court always goes to the team with the better record, so if Houston passes Utah, the official seeding number itself won't change, but it will have that effect since Houston will get home court. I think.

LRB
03-22-2007, 11:07 PM
Of course not. The model itself works just fine. But the model can't tell you things with much confidence as the data come so close together.

But the model knows very well that a team who, on average, nets a quarter-point per exchange of possessions will in fact kick the shit out of a team who on average gives up a quarter-point per exchange of possessions--in the long run, that is.



Chum, I think when taken on a league wide basis over a period of years, with 30 teams and over 82 games per team per year, the MOV has some level of accuracy. But I think it's outliers are vary much more than you make out. Take the Spurs and Mavs.

For example, the Spurs are averaging 0.17 more points in their Margin of Victory than the Mavs over 67 games. However the Mavs have won 9 more games than the Spurs and have won 2 out of 3 head to head meetings. In performance these two teams aren't even close. 9 games difference = over 13% of total games played.

Now take the spurs and the Suns. The Spurs have a MOV over the Suns of 1.07 ppg through 67 games. Yet the Suns have won 4 more games than the Spurs and 2 out of 3 head to head meetings. 4 games is almost 6% of the total games played.

Now let's look at the Bulls and the Heat. the Bulls have played 69 games to the heats 67 games. They've lost the same amount of games and the Bulls have won 2 more games. Yet the Bulls have 4.51 ppg edge in MOV.

MOV is all over the place. There are teams with huge MOV differences and little difference in records. There are teams with huge differences in records and little difference in MOV. There are teams with a better MOV, but a much worse record. There are teams with a better MOV and a better record. It's all over the freaking board. MOV is a poor rating factor because of how much 1 game can impact it. Houston won the other night by 50. Over an 82 game schedule that is a huge impact. I just don't buy MOV being a very reliable way of comparing two teams together.

chumdawg
03-22-2007, 11:47 PM
Chum, I think when taken on a league wide basis over a period of years, with 30 teams and over 82 games per team per year, the MOV has some level of accuracy. But I think it's outliers are vary much more than you make out. Take the Spurs and Mavs.

For example, the Spurs are averaging 0.17 more points in their Margin of Victory than the Mavs over 67 games. However the Mavs have won 9 more games than the Spurs and have won 2 out of 3 head to head meetings. In performance these two teams aren't even close. 9 games difference = over 13% of total games played.

Now take the spurs and the Suns. The Spurs have a MOV over the Suns of 1.07 ppg through 67 games. Yet the Suns have won 4 more games than the Spurs and 2 out of 3 head to head meetings. 4 games is almost 6% of the total games played.

Now let's look at the Bulls and the Heat. the Bulls have played 69 games to the heats 67 games. They've lost the same amount of games and the Bulls have won 2 more games. Yet the Bulls have 4.51 ppg edge in MOV.

MOV is all over the place. There are teams with huge MOV differences and little difference in records. There are teams with huge differences in records and little difference in MOV. There are teams with a better MOV, but a much worse record. There are teams with a better MOV and a better record. It's all over the freaking board. MOV is a poor rating factor because of how much 1 game can impact it. Houston won the other night by 50. Over an 82 game schedule that is a huge impact. I just don't buy MOV being a very reliable way of comparing two teams together.Those are all valid points, LRB. But to speak to your Houston example: how many teams have won a game by 50 this year? I'm guessing only Houston. Outliers are one thing, like the 50-point-game almost surely was. But what about a 25-point win? Is it ALWAYS an outlier? Well, very good teams will beat teams by 25 well more than once over a season.

Just look at each team's biggest win on the season. I doubt you will find a random element to it. I doubt you will find that the Celtics went off for 25+ on a team near as often as the Mavs did.

Don't let big numbers scare you away. They mean exactly what they mean. In other words, the "outliers" you describe may not be as meaningless as you might think they are.

And besides, the whole idea is that they all come out in the wash. If Boston up and wins a game by 30 (they won't) while at the same time losing most other games by ten, it won't make a lot of difference on their overall record. They are what their MOV says they are.

Yes, you should look at outliers with a careful eye. But in the course of a season, there are very, very few outliers anyway. Mostly, it is what it is.

Mavinator
03-23-2007, 12:03 AM
Y'all might have to forgive my partial bullheadedness on this subject. I am not a stats guy- I am a pure math guy with some limited background in statistics, and its my general opinion that predictive statistics are pretty close to useless when it comes to quantifying something like basketball, where the outcomes rely on more random factors than any equation devised by man could possibly hope to include. It's not really that I think Hollinger's model is bad as models go, its more that I think all statistical models are pretty flawed.

I'd say you're not making an honest attempt at ranking teams on goodness, but that you are deliberately ranking teams on badness while changing the meanings of words. That's not statistics. That's semantics. If he's changing the meaning of Margin of Victory, or of Strength of Schedule, or changing his formula in order for some teams to be ranked higher than others, then he'd be doing what you did. Of course, if he were playing semantics in order to come out with a ranking that everyone thought was wrong, well he would be an idiot. However, if he defined those things beforehand according to some method (that we'd all like to see), and has stuck to it, then yeah, his ranking is doing "pretty good." Whether or not it is sufficiently good is up to any individual reader to decide. But because it's an objective, statistical model, it can be tweaked in an honest effort to rank teams by goodness, and do an unarguably better job next time.

My point is probably exactly that there is no such thing as "the goodness of a team," at least not in the mathematical sense. Mavs > Suns, Suns > Mavs, that stuff doesn't exist- its a matter of opinion and not numbers. There's your regular season win-loss record and there's whether you win the championship or not- those are numbers, but i guarantee MOV is not the best historical stat for predicting these (my wager would be that, say, win-loss record is the best). I suppose you could attempt to quantify "likeliness to win a game against an average opponent on even terms" as "goodness," but I'd rather spend my day building a staircase to the moon than formulating a model to predict something like that. Rank the Suns number 1 due to margin of victory and what happens if on gameday Marion tears his ACL, or the opposing team's head player says something nasty in the papers about Stoudemire's mamma, or maybe even if Nash just has a really bad day? Basketball is decided by people, not a series of coinflips, and any attempt to quantify the pretty much infinite amount of randomness in a basketball game using five or six of your favorite statistical averages is precisely what I said it was- an arbitrarily chosen model.


I doubt his numbers are truly arbitrary, that he just pulled his model out of thin air, or that he didn't test anything. It is very easy to set an objective measure of goodness like win-loss, or playoff seeding, or playoff success, or head to head matchup, or whatever, and then to build a formula that will best fit past data. It's quite a bit harder to build one that will successfully predict future data. But because it's so easy to fit a model to past data, I would bet that he's done that. He just hasn't let us know what that past data is. My guess is he's being protective. If someone knows what data he's working on, then it makes it easier for them to do better than he did.

My question would be, how does he measure "success"? Its been demonstrated fairly admirably by several posters on this board and silly Mike Fisher that there doesn't appear to be any strong correlation between winning the title and being first in MOV. Maybe you are more likely to win the title if your MOV is above average but that seems trivial- no doubt you are more likely to win if you get a lot of rebounds or a lot of assists, also. My guess would be exactly that he devised his model based on what seemed to basically make sense to him. Of course you will win on average more games if you score more points than your opponents and if you tend to regularly defeat teams with good win-loss records- good job Hollinger, how do we possibly prove that your model isn't based on real empirical evidence? Of course, that's just me being skeptical of lazy sportswriters, but if John isn't willing to provide any evidence that he's done any research, I don't see any reason to believe his model is accurate to any great degree. (And on a deeper level, I believe he and the rest of the human race are incapable of creating an accurate model due to the nature of the sport.)



Also I agree strongly with pretty much everything LRB said- if we look at things from a intro stats point of view, in a sport with mind-boggling shitloads of random variables, almost none of which are independent from one another, and almost all of which have very high variances from game to game and season to season, it generally makes statistical predictive models pretty darn inaccurate

chumdawg
03-23-2007, 12:13 AM
With due respect, Mavinator, are you suggesting we pick 16 team names out of a hat and run the tournament from there?

You seem opposed to any measure--including win/loss--that qualifies a team.

Mavinator
03-23-2007, 01:49 AM
Haha basically. Stats are fun but somewhat masturbatory to sports fandom and basically inappropriate for something as grandiose as full-scale power rankings. Watch the games and that'll give you a better idea of which teams are good than any equation will




Modest revision upon reflection: If I HAD to base power rankings on numbers alone, I would do it by win-loss record, but no I don't think win-loss record is in any way a consistently accurate predictor of who will win a given game

chumdawg
03-23-2007, 02:18 AM
What do you think is then, if anything?

My goodness. Like I said. Pull the names from a hat, and go get 'em, is what you're saying, right?

LRB
03-23-2007, 03:02 AM
Those are all valid points, LRB. But to speak to your Houston example: how many teams have won a game by 50 this year? I'm guessing only Houston. Outliers are one thing, like the 50-point-game almost surely was. But what about a 25-point win? Is it ALWAYS an outlier? Well, very good teams will beat teams by 25 well more than once over a season.

Just look at each team's biggest win on the season. I doubt you will find a random element to it. I doubt you will find that the Celtics went off for 25+ on a team near as often as the Mavs did.

Don't let big numbers scare you away. They mean exactly what they mean. In other words, the "outliers" you describe may not be as meaningless as you might think they are.

And besides, the whole idea is that they all come out in the wash. If Boston up and wins a game by 30 (they won't) while at the same time losing most other games by ten, it won't make a lot of difference on their overall record. They are what their MOV says they are.

Yes, you should look at outliers with a careful eye. But in the course of a season, there are very, very few outliers anyway. Mostly, it is what it is.

A win doesn't have to be by 50. It the Celts played huge like Golden State on Friday and the Mavs laid an egg, a win by 30 is certainly a legitimate possibility although not a probability. Let's say that Dirk gets 2 quit tech's in the 1st half and Jet sprains his ankle in the 1st minute and Buckner and George don't get cleared to play. That would take Boston's MOV from it's current -3.53 to -3.04. That's a 13.9% change in MOV involving only 1.4% of games played. Now this doesn't accurately show how likely the Celts are to win their next game. The odds of multiple teams being without 2 of their top 3 players and 40% of their regular rotation more than once in a season get exponentially smaller than it happening once.

Yes given enough time, games, and teams this evens out. But it doesn't even out for a large % of teams each year. So it's a very inaccurate indicator when used to predict wins. Say maybe 60 to 65 % of the time it might be right when spread over all teams. But for a single team it could easily be wrong 60 to 65 % of the time.

Personally when judging the top teams, I think winning % is much better indicator. That's not a perfect predictor either, but it's better IMO. The best methods are probably the ones used by Vegas odds makers, but their flawed as well. Otherwise no one would bet.

I don't think that any Mathmatical formula doesn't have huge blind spots. It may be a good place to start, but you have to investigate and use different adjustments due to variables which can't be accurately represented in a formula. Even then it's only a best guess, and the chimp in the zoo might be better at picking the winning team by pure luck. In the NBA, on any given night any team can beat any other team.

I hate the BCS because of dumbass forumlas like this. The best way to decide which team is best is play the damn game. Formulas based on stats are never ever black and white when predicting future outcomes for NBA games. The best they can hope for is to be right more times than not. But rest assured, they will be wrong a significant number of times no matter what. Hollinger pisses me off with his dumbass defense of his formular as being so accurate as that it closely approaches 100% accuracy in future predictions when nothing could be further from the truth.

Dirkadirkastan
03-23-2007, 03:18 AM
MOV is a self-fulfilling prophecy. If you don't get the W-L record you expect, you say the W-L record doesn't mean anything.

Usually Lurkin
03-23-2007, 06:19 AM
Y'all might have to forgive my partial bullheadedness on this subject. I am not a stats guy- I am a pure math guy with some limited background in statistics, and its my general opinion that predictive statistics are pretty close to useless when it comes to quantifying something like basketball, where the outcomes rely on more random factors than any equation devised by man could possibly hope to include. It's not really that I think Hollinger's model is bad as models go, its more that I think all statistical models are pretty flawed.
oh, ok. You're just an academic snob. That makes sense. Most pure math folks are. ;) I bet your doctor loves it when he says, "Mavinator, there's a 5% chance that your surgeon will screw this up and accidentally kill you . . . " and you say, "I think statistics are bogus, you can't predict individual cases. Give me my slide rule...uh, FLAY-ven!"

I suppose you could attempt to quantify "likeliness to win a game against an average opponent on even terms" as "goodness," but I'd rather spend my day building a staircase to the moon than formulating a model to predict something like that.
I think there are a lot of people building that staircase. Or Space Elevator, at least. Sounds like a dream job to me. I would have a hard time deciding between that and Statistical Analyst for the Dallas Mavericks. Wait, no I wouldn't. Mavs rule.

Rank the Suns number 1 due to margin of victory and what happens if on gameday Marion tears his ACL, or the opposing team's head player says something nasty in the papers about Stoudemire's mamma, or maybe even if Nash just has a really bad day?
The exact thing that happens with opinion models, the person in charge of making some decision gets to decide whether the assumptions of the original model have been sufficiently violated or not.

Basketball is decided by people, not a series of coinflips, and any attempt to quantify the pretty much infinite amount of randomness in a basketball game using five or six of your favorite statistical averages is precisely what I said it was- an arbitrarily chosen model.
It's not arbitrary. That would be pulling stuff out of a hat. Hollinger's model is "relatively" good. Whether "relatively" is good enough is up to the reader. I know, I know, the old joke is that the difference between mathematicians and an engineer (statistician in this case) is that the mathematician doesn't know the term "good enough." But you get the point. "Good enough" is an approximation function.

People make millions and billions of dollars with statistical formulations of the stock market, which is easily as random as basketball.

My question would be, how does he measure "success"?
I think that's everybody's question.

Usually Lurkin
03-23-2007, 06:45 AM
Hollinger's formula was pretty worthless in predicting that Miami would go on the role it did after Dshoulder went down with injury. In truthfulness I don't see how it's very predictive of what will happen in the future at all.
How many opinion based polls predicted that?

As for outliers like Houston's 50 point win, there are a number of ways to deal with them. A lot of people have suggested capping any margin of victory at 15 points or something. This is basically a "big win is a big win" approach. Another is to ignore very huge values, or factor them in with a little less value than you might otherwise. This is a "well, it happened, but it's a little suspicious" approach. More commonly in statistical analysis, I think, is to multiply all the values by some fractional value (or log value), so that the difference between margins of 5 and 10 and margins between 20 and 50 isn't quite so big. Another common approach is to change the metric by which values in the formula are changed to approximate whatever is defined as "goodness."

I don't think that any Mathmatical formula doesn't have huge blind spots. It may be a good place to start, but you have to investigate and use different adjustments due to variables which can't be accurately represented in a formula. Even then it's only a best guess, and the chimp in the zoo might be better at picking the winning team by pure luck. In the NBA, on any given night any team can beat any other team.

I think all opinions are based on a bunch of pseudo-statistical judgements. It's just that usually, we just have the vaguest notions going through our heads, like, "well, the mavs are winning more games than their opponent, but they win more easily when their best defensive players are healthy, and they're not, so I'll factor that in, and usually, people don't perform as well when they're tired, and Dirk has been playing a lot of minutes, so I'll factor that in, and . . . "
The only thing that's different about a statistical approach is that it tries to put those judgements in written form.
The downside is that the more readable it is (the fewer judgements there are in it), the less accurate it will be.
On the upside, it can always be made more accurate, and with enough complexity, it will do exactly what the opinion based judgements do. It's also objective, so it won't change based on random factors of the judger. Human opinions are susceptible to events like like stubbing a toe on the way to work or getting ticked off because you've lost money on a team, or favoring teams because you are part owner or because you used to play for them (Kerr!).

LRB
03-23-2007, 01:04 PM
How many opinion based polls predicted that?

I never espoused pure opinion polls. But I'd bet that the vegas style system that combines opinion with objective stats, is much more accurate that a pure statistical model.


As for outliers like Houston's 50 point win, there are a number of ways to deal with them. A lot of people have suggested capping any margin of victory at 15 points or something. This is basically a "big win is a big win" approach. Another is to ignore very huge values, or factor them in with a little less value than you might otherwise. This is a "well, it happened, but it's a little suspicious" approach. More commonly in statistical analysis, I think, is to multiply all the values by some fractional value (or log value), so that the difference between margins of 5 and 10 and margins between 20 and 50 isn't quite so big. Another common approach is to change the metric by which values in the formula are changed to approximate whatever is defined as "goodness."



Those methods help and increase the general accuracy, but there still holes that can be spotted by a human brain. A pure statistical model cannot factor in unknown future changes. So if something changes that you didn't have statistical evidence on when you created your formula, you're left out in the cold. That's why you need human input to tweak the forumula or adjust the data for unaccounted for variables that come into play to get the best possible prediction. But even the best possible prediciton still isn't going to be near 100% accurate.


I think all opinions are based on a bunch of pseudo-statistical judgements. It's just that usually, we just have the vaguest notions going through our heads, like, "well, the mavs are winning more games than their opponent, but they win more easily when their best defensive players are healthy, and they're not, so I'll factor that in, and usually, people don't perform as well when they're tired, and Dirk has been playing a lot of minutes, so I'll factor that in, and . . . "
The only thing that's different about a statistical approach is that it tries to put those judgements in written form.
The downside is that the more readable it is (the fewer judgements there are in it), the less accurate it will be.
On the upside, it can always be made more accurate, and with enough complexity, it will do exactly what the opinion based judgements do. It's also objective, so it won't change based on random factors of the judger. Human opinions are susceptible to events like like stubbing a toe on the way to work or getting ticked off because you've lost money on a team, or favoring teams because you are part owner or because you used to play for them (Kerr!).

It's impossible to have a totally objective formula, no matter how complex it is. The formula will be based upon the subjectivity of it's creator. There will be randomness built into a formula. A formula's achille's (sp) heal will always be it's inability to adapt to new data that wasn't present when it was created. For example, this year the NBA instituted some new rules. One new rule was giving refs more latitude on calling Tech's. No data was present before this season on how that new rule would affect outcomes. Now if you base predicitons on a combination of objective stats with subjective and flexible input from humans it can adapt better. Essentially you're tweaking your forumula as new data becomes present.

Now there are some keys to obtaining the human input. 1. Level of success must be measured and reported. Most opinion polls and formula's don't do this to my knowledge. 2. Humans must be held accountable for their input. Sure Kerr has built in biases. And he can speak out of his ass based on those biases all day and still have a job. The same is not true for Vegas odds makers. 3. Human input should be spread over many individuals instead of just a very few. 4. Projected results should be constructed in ranges instead of exact numbers with the expected likelyhood that results will fall withing that range. Then you can use the high, low, middle or some other point in the range for an exact number, but with informaiton on what the odds are that the results will be more or less than that number. I'm sure there are many more, but these 4 are a very good start.

I have a huge problem with formula's that are painted as being near 100% accurate in predicting future outcomes of NBA games, because I don't believe that's possible over any significant # of games. I also think that the NBA has too many new variables introduced each season for which there isn't sufficient data to do statisitcal analysis before the season, for a set in stone formula to be as accurate as one that is tweaked during the season as new data becomes available.

endrity
03-23-2007, 01:25 PM
Wow, the sports statistics is turning into a huge thread. I never realized there were this many math and stats guys in the forum. I myself study economics so I have some background. Anyway, I thought you guys might find these two sites interesting. They bring a new twist to basketball statistics, especially Berri's one (the first underneath)
http://dberri.wordpress.com
http://www.thesportseconomist.com/

Also, trueHoop has an article today about sport's statistics. The main point is that there still is a lot of debate whether there is an ultimate truth in sports that statistics can reveal or not.

Usually Lurkin
03-23-2007, 01:49 PM
I never espoused pure opinion polls. But I'd bet that the vegas style system that combines opinion with objective stats, is much more accurate that a pure statistical model. .
I don't think anyone, vegas included, predicted that Miami would do well when DWheels went down. It just sometimes sounds like you are measuring the use of statistics against the same impossibly high standard that Hollinger is claiming to have already reached.

Those methods help and increase the general accuracy, but there still holes that can be spotted by a human brain. A pure statistical model cannot factor in unknown future changes. So if something changes that you didn't have statistical evidence on when you created your formula, you're left out in the cold. That's why you need human input to tweak the forumula or adjust the data for unaccounted for variables that come into play to get the best possible prediction. But even the best possible prediciton still isn't going to be near 100% accurate.
This I pretty much agree with. Maybe I find it a bigger advantage than you do to have a method by which statistical predictions can become measurably more accurate. The objectivity of a statistical model lets you know exactly how reliable statistical factors are, or how much they can be factored into your judgement. If you have a way to make the statistical predictions better and better by degrees, then you have a way to make the statistical part of your overall judgement better and better by degrees.

I agree that Hollinger's fault lies in selling his stats too hard. I think he should have some error bars on his rankings, as in Spurs are 109.8, plus or minus .8; Mavs are 109.4 plus or minus .9, etc.

chumdawg
03-23-2007, 01:54 PM
This is an intriguing discussion. But....where did this idea come from that Hollinger was talking about his power ranking as some tool to predict series outcomes with?

Mavinator
03-23-2007, 02:00 PM
What do you think is then, if anything?

My goodness. Like I said. Pull the names from a hat, and go get 'em, is what you're saying, right?

Haha hardly... Games and playoff rounds won are concrete and they've already happened; statistical prediction of basketball is fuzzy math and often unlikely to happen. The teams that get to the playoffs and win do so because they've earned it and because they're good in an entirely unquantifiable way; my point is you're out of luck if you want to predict that accurately midseason by examining their rebounding differential or whatever...

oh, ok. You're just an academic snob. That makes sense. Most pure math folks are. I bet your doctor loves it when he says, "Mavinator, there's a 5% chance that your surgeon will screw this up and accidentally kill you . . . " and you say, "I think statistics are bogus, you can't predict individual cases. Give me my slide rule...uh, FLAY-ven!"

No I don't talk to doctors, I look down on applied sciences also. Haha just kidding. Actually one of my best friends in college (a general math major) did his thesis senior year doing regressions on baseball stats and trying to come up with a fairly accurate predictor of who would win the world series any given year. Anyway the guy worked for months trying different combinations of like 40 different baseball stats, and came out with an 80 page thesis (I guarantee a lot more work than Hollinger did) by the end of the year with a lot of interesting attempts, but basically his conclusion was that selecting different stats produced wildly different results and that no selection seemed overwhelmingly accurate. He's a colossal baseball fan and when he gave his presentation at the end of the year, he said baseball is decided by the intangibles of the players and not by an assembly of random variables- and that's in baseball, which is think is a somewhat more statistically quantifiable game than basketball.

(Admittedly, at the end of the year, most of the professorial criticism was along the lines of "Your thesis wasn't about math, it was about baseball"- but that's the way it should be, right?)

I think there are a lot of people building that staircase. Or Space Elevator, at least. Sounds like a dream job to me. I would have a hard time deciding between that and Statistical Analyst for the Dallas Mavericks. Wait, no I wouldn't. Mavs rule.

Hey man, I sell blood for a living, I couldn't afford the first elevator cable. You work on that contraption, I'll stick to my staircase

The exact thing that happens with opinion models, the person in charge of making some decision gets to decide whether the assumptions of the original model have been sufficiently violated or not.

But what good is a predictive stat that has to be updated on a game to game basis?

It's not arbitrary. That would be pulling stuff out of a hat. Hollinger's model is "relatively" good. Whether "relatively" is good enough is up to the reader. I know, I know, the old joke is that the difference between mathematicians and an engineer (statistician in this case) is that the mathematician doesn't know the term "good enough." But you get the point. "Good enough" is an approximation function.

Well, ok, the truth is I think we're both right here. I mean, we're trying to predict who's going to win the title here. I pull a name out of a hat and I have a one in thirty chance of being right- you use the hollinger model and have maybe a one in eight chance of being right, or one in six or one in four. There's a definite improvement there, right? But even a statistician would admit that its not really statistically significant- a 15%-odd chance of winning the title isn't nearly enough to draw any realistic conclusions. Same on a game-to-game basis- determining you have maybe a 65% chance to win a given game doesn't really do you any good (unless, I suppose, you're a sports gambler, but that's a whole different ballgame). What a statistician MIGHT be able to construct is a set of teams that are likely to win it- say, he could construct a model that says with 95% confidence either Dallas, PHX, SA, Houston, Detroit, or Utah is going to win it all, and then we have something that's statistically significant, but what good does it do us fans? We could tell that by looking. (And, one again, I expect we could construct a more accurate model along these lines using W-L than MOV-whatsis)

I think that's everybody's question.

Haha then we agree! Actually I heard Hollinger's going to publish an article this week announcing that he has finished building his moon staircase. I guess we're not going to be allowed to see it or informed how he built it, but he has confirmed that the moon is indeed made of cheese

This has been a really fun talk, its been a long time since I got sucked into a DM discussion like this, normally there's so little to talk about when we win like, every game. But I'm gonna have to drop out soon, I have to get back to triangulating topological spaces and shit while other people do practical things... I fully throw my support behind LRB, I agree with everything he says

alexamenos
03-23-2007, 02:25 PM
Actually one of my best friends in college (a general math major) did his thesis senior year doing regressions on baseball stats and trying to come up with a fairly accurate predictor of who would win the world series any given year. Anyway the guy worked for months trying different combinations of like 40 different baseball stats, and came out with an 80 page thesis (I guarantee a lot more work than Hollinger did) by the end of the year with a lot of interesting attempts, but basically his conclusion was that selecting different stats produced wildly different results and that no selection seemed overwhelmingly accurate. He's a colossal baseball fan and when he gave his presentation at the end of the year, he said baseball is decided by the intangibles of the players and not by an assembly of random variables- and that's in baseball, which is think is a somewhat more statistically quantifiable game than basketball.

I had a grad stats class which was devoted to developing a multiple regression model. The profs wanted me to do some lame macro-economic stuff, but I went with the ponies.

Basically....after every race, a horse is given a rating in lieu of a time. This rating (i forget what it's called, now) normalizes the time so that horses running on different tracks in different conditions can be compared to one another. So, I looked at lots and lots and lots of past races, and developed a model that would predict a horse's rating for his next race. The model included things like how good a jockey was riding the horse, the horses past performance in longer, shorter, and equi-distant races, whether the horse had recently started taking dopamine, all kinds of stuff....

Statistically speaking, the model was nails. R-sq > 80%, off-the chart significance amongst all the coefficients, etc., etc.... the profs were shocked, shocked i tell you, that i was able to get anything at all. they made personal copies of the model for their coming trips to the tracks. they wrote songs about my mathematical prowess and sent me flowers.

but....

here's the typical result....you plug in all the data on the horses for a race, and you get predicted ratings kind of like this....

Horse A - 110
Horse B - 108
Horse C - 106

then you take a look at the odds and you'd see this:

Horse A - 2-1
Horse B - 5-2
Horse C - 3-1

IOW, for all of the robustness of the model, it didn't tell you a damn thing worth knowing. In those rare circumstances where a longer-odds horse would get a higher rating than the lower odds horse, it was invariably because the betting public knew something the model didn't, and not the other way around.

which points to, i s'pose, my take on statistical models....they can be useful bits of information, but nothing more.

cheers

Five-ofan
03-23-2007, 02:58 PM
I had a grad stats class which was devoted to developing a multiple regression model. The profs wanted me to do some lame macro-economic stuff, but I went with the ponies.

Basically....after every race, a horse is given a rating in lieu of a time. This rating (i forget what it's called, now) normalizes the time so that horses running on different tracks in different conditions can be compared to one another. So, I looked at lots and lots and lots of past races, and developed a model that would predict a horse's rating for his next race. The model included things like how good a jockey was riding the horse, the horses past performance in longer, shorter, and equi-distant races, whether the horse had recently started taking dopamine, all kinds of stuff....

Statistically speaking, the model was nails. R-sq > 80%, off-the chart significance amongst all the coefficients, etc., etc.... the profs were shocked, shocked i tell you, that i was able to get anything at all. they made personal copies of the model for their coming trips to the tracks. they wrote songs about my mathematical prowess and sent me flowers.

but....

here's the typical result....you plug in all the data on the horses for a race, and you get predicted ratings kind of like this....

Horse A - 110
Horse B - 108
Horse C - 106

then you take a look at the odds and you'd see this:

Horse A - 2-1
Horse B - 5-2
Horse C - 3-1

IOW, for all of the robustness of the model, it didn't tell you a damn thing worth knowing. In those rare circumstances where a longer-odds horse would get a higher rating than the lower odds horse, it was invariably because the betting public knew something the model didn't, and not the other way around.

which points to, i s'pose, my take on statistical models....they can be useful bits of information, but nothing more.

cheers
the rating is called a speed index.

alexamenos
03-23-2007, 03:02 PM
the rating is called a speed index.

i knew it was some fancy schmancy term....

Mavinator
03-23-2007, 03:14 PM
Haha that's really neat... nice work alexamenos.

From a theoretical viewpoint, isn't odds-making less about who is going to win and more about who people think is going to win? That is, aren't they trying to predict how people will bet rather than how the game/race/whatever will play out?

I'm not an oddsmaker thankfully but thats just how it seems to me

LRB
03-23-2007, 03:49 PM
I don't think anyone, vegas included, predicted that Miami would do well when DWheels went down. It just sometimes sounds like you are measuring the use of statistics against the same impossibly high standard that Hollinger is claiming to have already reached.


I seriously doubted that they (Vegas odds makers) got it right for the 1st few games. But I bet they corrected sooner and are more in line right now than Hollinger's overrated formula or pretty much any other static formula. But I was using that argument more to show that Hollinger's model comes nowhere close to 100% accuracy. Vegas doesn't either, but I believe it comes much close than others.


I agree that Hollinger's fault lies in selling his stats too hard. I think he should have some error bars on his rankings, as in Spurs are 109.8, plus or minus .8; Mavs are 109.4 plus or minus .9, etc.

I would love to see something like this added.

alexamenos
03-23-2007, 04:47 PM
From a theoretical viewpoint, isn't odds-making less about who is going to win and more about who people think is going to win?

the initial line is the odds-makers guess as to how people will bet....as time goes on, the line reflects how people are actually betting.

chumdawg
03-23-2007, 05:18 PM
the rating is called a speed index.And more specifically, he was probably talking about Beyer speed figures.

pkdumas
03-23-2007, 10:07 PM
I think the Mavs just passed the Spurs in the "Hollinger Power Rankings" tonite. Mavs win by double-digits on the road, Spurs win by 1 at home. We were only .02 behind before these games. I kind of liked not being the best. But we'll see in the morning if this holds up.

Usually Lurkin
03-24-2007, 05:22 AM
No such luck, pkdumas.

Remember last week when the 2nd ranked Mavs were ahead of the 1st ranked Spurs on every category except MOV (and the hidden home-away)? Here's today:

SPURS
HOLLINGER RANK RATING RECORD MARGIN SOS REC (L10) MAR (L25%) SOS (L25%) STREAK DIV CONF
1 1111 109.345 48-20 +8.18 .507 8-2 +11.94 .483 Won 2 2nd 3rd
Prev Rnk: 1 | Prev Rat: 109.269 | Lst Game: DET, W 90-89 | Nxt 3: @SEA (Sun), @GS

MAVS
HOLLINGER RANK RATING RECORD MARGIN SOS REC (L10) MAR (L25%) SOS (L25%) STREAK DIV CONF
2 2222 109.105 57-11 +8.18 .508 8-2 +10.24 .506 Won 5 1st 1st
Prev Rnk: 2 | Prev Rat: 109.183 | Lst Game: @BOS, W 109-95 | Nxt 3: @ATL (Sun), @NOK (Tues), MIL (Wed)

Here's a question for those of you digging into the formula. If the home court advantage is 3 points, then shouldn't the Spurs' 1 point victory over the Pistons be a -2 MOV? How did their rating go up? Isn't that how the Mavs moved down a few weeks ago? Also, the Mavs won by 14 points (on the road no less). How the heck did their rating drop?

mary
03-24-2007, 07:47 AM
Here's a question for those of you digging into the formula. If the home court advantage is 3 points, then shouldn't the Spurs' 1 point victory over the Pistons be a -2 MOV? How did their rating go up? Isn't that how the Mavs moved down a few weeks ago? Also, the Mavs won by 14 points (on the road no less). How the heck did their rating drop?

I don't have alot of time, but just off the top of my head the change in SOS would've heavily favored the Spurs last night. Also, the game to fall out of the Mavs L25 would've been a 106-89 victory over the sixers, having a negative impact for MOV-L25 and it looks like the game to fall out of the L25 for the Spurs would've been a loss to the Heat of 85-100, having a postive impact on their MOV-L25.

Maybe Chum can verify this.

Usually Lurkin
03-24-2007, 08:19 AM
I don't have alot of time, but just off the top of my head the change in SOS would've heavily favored the Spurs last night.

Maybe Chum can verify this.
are hollinger's the mis-reported data that folks was talking about earlier?
Is a +11 against losing teams (.483) in the last 25% really so much better than a +10 against winning teams (.506) that it outweighs all the other factors?

dude1394
03-24-2007, 11:21 AM
my goodness there is a lot of interest in trying to decipher what hollinger is doing. Lot's o' stat majors it seems.

Where do all of the statistics come from? Not engineering, is it business or computer science?

wmbwinn
03-24-2007, 11:47 AM
I still like these stats better:

http://www.dougstats.com/06-07Teams.html

The Mavs have significantly improved in shooting percentage and 3 point percentage. Their improvements compared to earlier rankings have significantly improved. They were in the middle of the bunch or average in shooting percentage.

The Mavs are still the only team in the top ten for both defense and offense.

The Mavs are only outrebounded by the Jazz as to rebound differential.

The Mavs still lead the league in points scored versus allowed differential.

Our field goal percentage differential is WHOLE lot better, we are now ranked number 5 in the league and used to be in the middle of the pack just a month ago.

Defensive field goal percentage is better. 3 point field goal percentage remarkably improved.

We are now #3 on turnovers (meaning only two teams commit fewer turnovers). We still are near the top in assist differential.

We have been committing more fouls and have slipped downward there some.

There is still only one team that allows its opponent to shoot fewer shots for the season against them than the Mavs. That better ranked team changed though. Last time I checked, the Pistons did the best defensive job of preventing shots in the first place. Now, the most stingy team is Utah. http://www.dougstats.com/06-07RD.Team.Opp.txt

The Mavs are now the number one team as how many total shots for the season have been made against them. It is basically a tie with the Jazz and Pistons.

So, the Mavs have the defense of the Jazz and Pistons but have a far superior offense...

Anyway, the main stat is the wins and losses... #1

More important stat is who is holding the trophy at the end...

LRB
03-25-2007, 12:14 PM
Hopefully Avery won't make any more stupid moves today like not playing Dirk the entire 4th quarter like he did against Boston. That really hurt our chances to win a championship according to Hollinger. If Dirk had played the entire 4th quarter and the mavs could have pushed to lead to 30 or so, we might have had a chance of beating the Spurs in the playoffs. Now sadly we will lose according to Hollinger. You don't win championships beating suckass teams in late March by low double digits so you can give your star players rest.

mary
03-30-2007, 10:43 AM
The Hollinger Rankings are about to be overtaken by the Texas Triangle.

Something about that is kinda cool...

Flacolaco
03-30-2007, 10:48 AM
The Hollinger Rankings are about to be overtaken by the Texas Triangle.

Something about that is kinda cool...

not if you think about the possibility of the Mavericks likely playing both of those teams in the playoffs. :(

Dirkadirkastan
03-31-2007, 01:19 AM
not if you think about the possibility of the Mavericks likely playing both of those teams in the playoffs. :(

omg the mavs will actually have to beat someone GOOD to win the title?? :eek:

Looks like someone wants the trophy minus the glory. I prefer to adopt the attitude that if the Mavs had to play 29 seven-game series against every team in the NBA, they'd win every single one of them. Let's get a little confidence in here, jeez...

MartinGT
03-31-2007, 04:16 AM
LIVE WITH IT!

Thespiralgoeson
03-31-2007, 04:27 AM
LIVE WITH IT!

Shut up.