View previous topic :: View next topic |
Author |
Message |
Eli W
Joined: 01 Feb 2005 Posts: 348
|
Posted: Tue Mar 13, 2007 3:19 pm Post subject: Current season Win Scores/Wins Produced |
|
|
The Wages of Wins blog linked to a new site that calculates players' Win Scores and Wins Produced for the current season.
Here's the blog post introducing the site:
http://winscore.blogspot.com/2007/03/finally-something-to-play-with.html
And the site itself:
http://www.jasonchandler.com/basketball/
It appears that there is a separate stats community on the web growing out of the Wages of Wins that unfortunately seems disconnected from APBRmetrics. I know lots of people here have various objections to what Berri et al are doing (and vice versa), but more integration would be nice IMO. _________________ Eli W. (formerly John Quincy)
CountTheBasket.com |
|
Back to top |
|
|
kjb
Joined: 03 Jan 2005 Posts: 725 Location: Washington, DC
|
Posted: Tue Mar 13, 2007 3:51 pm Post subject: |
|
|
My understanding of Berri's claims about this formula is that I should be able to take the per minute Wins Produced scores, multiply by the total minutes played by each player, and then add those Total Wins Produced scores together to get an accurate estimate of each team's total wins. I've gone through a handful of teams and found some teams that are very close (WP puts Boston at 25.6 wins so far, and the Celts are at 25). But I've also found large differences (WP puts Golden State at 45.2 wins -- Warriors are at 30). _________________ My blog |
|
Back to top |
|
|
gabefarkas
Joined: 31 Dec 2004 Posts: 961 Location: Durham, NC
|
Posted: Tue Mar 13, 2007 4:19 pm Post subject: Re: Current season Win Scores/Wins Produced |
|
|
John Quincy wrote: |
It appears that there is a separate stats community on the web growing out of the Wages of Wins that unfortunately seems disconnected from APBRmetrics. I know lots of people here have various objections to what Berri et al are doing (and vice versa), but more integration would be nice IMO. |
Weird...they're like the bizarro us, worshiping the red sun of Berri. |
|
Back to top |
|
|
hpanic7342
Joined: 16 Feb 2005 Posts: 201
|
Posted: Tue Mar 13, 2007 5:19 pm Post subject: |
|
|
Apparently David Lee adds the most wins per minute.
Yeah. |
|
Back to top |
|
|
Dan Rosenbaum
Joined: 03 Jan 2005 Posts: 537 Location: Greensboro, North Carolina
|
Posted: Tue Mar 13, 2007 5:52 pm Post subject: |
|
|
WizardsKev wrote: | My understanding of Berri's claims about this formula is that I should be able to take the per minute Wins Produced scores, multiply by the total minutes played by each player, and then add those Total Wins Produced scores together to get an accurate estimate of each team's total wins. I've gone through a handful of teams and found some teams that are very close (WP puts Boston at 25.6 wins so far, and the Celts are at 25). But I've also found large differences (WP puts Golden State at 45.2 wins -- Warriors are at 30). |
I think Wins Produced for a team adds up to the efficiency differential rather than wins. Berri just expresses it in terms of wins to make it easier to understand. Also, I think you need to account for trades when you do this adding up, so a team like Golden State is liable to be off a lot. |
|
Back to top |
|
|
kjb
Joined: 03 Jan 2005 Posts: 725 Location: Washington, DC
|
Posted: Tue Mar 13, 2007 6:58 pm Post subject: |
|
|
I thought trades might be an issue, but a quick eyeballing of Golden State's roster on that page showed Al Harrington's minutes in the 900s, and had Troy Murphy only for the minutes he played in Golden State.
I'm not sure how the numbers can add up to efficiency differential. I didn't spend a lot of time looking at it, though. _________________ My blog |
|
Back to top |
|
|
Dan Rosenbaum
Joined: 03 Jan 2005 Posts: 537 Location: Greensboro, North Carolina
|
Posted: Tue Mar 13, 2007 8:22 pm Post subject: |
|
|
WizardsKev wrote: | I thought trades might be an issue, but a quick eyeballing of Golden State's roster on that page showed Al Harrington's minutes in the 900s, and had Troy Murphy only for the minutes he played in Golden State.
I'm not sure how the numbers can add up to efficiency differential. I didn't spend a lot of time looking at it, though. |
And, of course, it is possible that this site is not calculating Wins Produced correctly. I don't Berri has ever publicly revealed the actual formula, and I think the approximation that Berri has offered is not going to work well with a high-pace, poor defensive team like Golden State. |
|
Back to top |
|
|
Mike G
Joined: 14 Jan 2005 Posts: 1772 Location: Delphi, Indiana
|
Posted: Wed Mar 14, 2007 7:07 am Post subject: |
|
|
WizardsKev wrote: | My understanding of Berri's claims about this formula is that I should be able to take the per minute Wins Produced scores, multiply by the total minutes played by each player, and then add those Total Wins Produced scores together to get an accurate estimate of each team's total wins. I've gone through a handful of teams and found some teams that are very close (WP puts Boston at 25.6 wins so far, and the Celts are at 25). But I've also found large differences (WP puts Golden State at 45.2 wins -- Warriors are at 30). |
Should stat-based 'player wins' add up to actual team wins or to pythagorean expectation? I would think the latter. But then, GS has won about 1 game more than their PythEx; so that makes it worse, not better.
The formula (if that's all there is to it) looks like one of many that reward offense and (almost) disregard defense. (Pts+Reb+etc)/Min ?? Come on. Is that what this excitement is really about?
As for speculation about a Berri camp growing parallel to APBRMetrics: If he were collaborating with us, could he call his ideas original?
I haven't said much about the Berri phenomenon, nor have I looked into it. But a red flag is in the subtitle of the book: "Taking Measure of the Many Myths .." Isn't 'mythbusting' ever going to run its course? Is there an eternal supply of dweebs thinking some formulaic 'magic wand' is going to give them insights unattainable by everyone else?
I looked at the Pistons page, and see Dale Davis contributes more than Rasheed, Rip, Prince, etc. Am I reading that right? _________________ 40% of all statistics are wrong. |
|
Back to top |
|
|
kjb
Joined: 03 Jan 2005 Posts: 725 Location: Washington, DC
|
Posted: Wed Mar 14, 2007 8:46 am Post subject: |
|
|
Mike G wrote: | WizardsKev wrote: | My understanding of Berri's claims about this formula is that I should be able to take the per minute Wins Produced scores, multiply by the total minutes played by each player, and then add those Total Wins Produced scores together to get an accurate estimate of each team's total wins. I've gone through a handful of teams and found some teams that are very close (WP puts Boston at 25.6 wins so far, and the Celts are at 25). But I've also found large differences (WP puts Golden State at 45.2 wins -- Warriors are at 30). |
Should stat-based 'player wins' add up to actual team wins or to pythagorean expectation? I would think the latter. But then, GS has won about 1 game more than their PythEx; so that makes it worse, not better. |
Maybe it is supposed to be expected wins. Golden State wasn't the only significant difference, though. I've tossed the spreadsheet I was working from, but I did run the team I know best. The Wizards' pythagorean expected # of wins is 31 so far this season. They've actually won 34 so far. The Berri attributed calculations from this site add up to 41.4.
Quote: | I looked at the Pistons page, and see Dale Davis contributes more than Rasheed, Rip, Prince, etc. Am I reading that right? |
According to the calculations on this site, Dale Davis, Carlos Delfino and Nazr Mohammed contribute more per minute than Tayshaun Prince, Antonio McDyess, Richard Hamilton, and Rasheed Wallace. _________________ My blog |
|
Back to top |
|
|
Dan Rosenbaum
Joined: 03 Jan 2005 Posts: 537 Location: Greensboro, North Carolina
|
Posted: Wed Mar 14, 2007 9:15 am Post subject: |
|
|
This WP48 approximation is not including a team adjustment, so things like pace or defense that is not accounted for with blocks, steals, defensive rebounds, or personal fouls will not be in there. So a high-paced team who is poor on defense but gets lots of steals is going to have more WP48 wins than actual wins. A slow-paced team who is good on defense but does not get a lot of steals should have a lot fewer WP48 wins.
And it will be interesting when folks actually go through and see how well this WP48 without a team adjustment explains team wins. It won't be anywhere near the 95% of Wins Produced with a team adjustment and then Berri will have to explain why this team adjustment that "isn't important" actually is important.
Last edited by Dan Rosenbaum on Wed Mar 14, 2007 9:25 am; edited 2 times in total |
|
Back to top |
|
|
Mike G
Joined: 14 Jan 2005 Posts: 1772 Location: Delphi, Indiana
|
Posted: Wed Mar 14, 2007 9:22 am Post subject: |
|
|
Yeah, I'm calculating the Wiz PW = 30.5
And I've got their eW = 29.8. My standard error is about 1.5 W/team -- but I haven't split seasons for traded players like Iverson and Webber, so those teams are off by as much as 3.8, from PW.
With the GS-Ind trade, eW shows MurphLeavy = HarringJax. But Diogu gives Indy the edge. _________________ 40% of all statistics are wrong. |
|
Back to top |
|
|
Mike G
Joined: 14 Jan 2005 Posts: 1772 Location: Delphi, Indiana
|
Posted: Wed Mar 14, 2007 9:25 am Post subject: |
|
|
deleted duplicate _________________ 40% of all statistics are wrong. |
|
Back to top |
|
|
Eli W
Joined: 01 Feb 2005 Posts: 348
|
Posted: Wed Mar 14, 2007 9:32 am Post subject: |
|
|
Dan Rosenbaum wrote: | And it will be interesting when folks actually go through and see how well this WP48 without a team adjustment explains team wins. It won't be anywhere near the 95% of Wins Produced with a team adjustment and then Berri will have to explain why this team adjustment that "isn't important" actually is important. |
Berri argues against that in this post, though he just gives the correlation between position-adjusted Win Score per minute and WP48 with a team adjustment (0.994), rather than between position-adjusted Win Score per minute and team wins. _________________ Eli W. (formerly John Quincy)
CountTheBasket.com |
|
Back to top |
|
|
Dan Rosenbaum
Joined: 03 Jan 2005 Posts: 537 Location: Greensboro, North Carolina
|
Posted: Wed Mar 14, 2007 9:48 am Post subject: |
|
|
John Quincy wrote: | Dan Rosenbaum wrote: | And it will be interesting when folks actually go through and see how well this WP48 without a team adjustment explains team wins. It won't be anywhere near the 95% of Wins Produced with a team adjustment and then Berri will have to explain why this team adjustment that "isn't important" actually is important. |
Berri argues against that in this [url=http://dberri.wordpress.com/2007/02/25/introducing-pawsmin-–-and-a-defense-of-box-score-statistics/]post[/url], though he just gives the correlation between position-adjusted Win Score per minute and WP48 with a team adjustment (0.994), rather than between position-adjusted Win Score per minute and team wins. |
I completely agree that at the individual-level Wins Produced with and without a team adjustment are highly correlated. I said that last summer. But it is still the case that Wins Produced without a team adjustment will explain far less than 95% of team wins. This also was something I pointed out last summer (and something I have pointed out to Berri in e-mails), so it is hard to believe that Berri didn't know this. |
|
Back to top |
|
|
Eli W
Joined: 01 Feb 2005 Posts: 348
|
Posted: Wed Mar 14, 2007 9:53 am Post subject: |
|
|
Interesting. So you think this is likely a case where the correlations aren't transitive. _________________ Eli W. (formerly John Quincy)
CountTheBasket.com |
|
Back to top |
|
|
|