APBRmetrics

Posted: **Fri Jun 24, 2011 12:30 pm**

NBA Player of the Year might be a satisfactory designation combining contributions of the regular season and the postseason. And of course, there will be runners-up; and in fact a whole new ranking.

I'll do it with eWins and an addendum. Just adding equivalent wins from the season and the playoffs, there's no acknowledgment or bonus for winning it all. The playoffs end after 4 rounds -- both teams 'go home for the summer'. But what if it went beyond 4 rounds? -- as if there were no limit to the playoffs, even.

A playoff series goes 4 to 7 games; avg might be 5.5 . If the NBA champ should go another series, they can expect another 5.5 games; and if they win that series (at 50% likelihood), another 5.5, and so on. This sequence predicts an avg 11 additional games from the winner of a series, in an endless tournament.

For the Mavs' I'm multiplying their playoff eWins by 31/21 -- equivalent to granting them 10 games' eWins on top of their 21 playoff games. Now the playoff 'equivalent equivalent wins' leaders look like this:

Code: Select all

e2W      per36 rates    tm   Min   Eff%    Sco   Reb   Ast    PF   Stl   TO   Blk
6.14   Dirk Nowitzki   Dal   826   .607   31.6   8.6   2.8   2.0    .5   2.9   .6
4.12   Dwyane Wade     Mia   824   .563   27.6   7.5   5.0   2.4   1.5   2.9  1.2
4.06   Lebron James    Mia   923   .552   22.6   7.8   6.1   2.2   1.4   2.6  1.0
3.04   Derrick Rose    Chi   651   .496   27.5   4.4   7.6   1.9   1.2   3.3   .6
2.94   Kevin Durant    Okl   724   .571   25.5   7.4   2.6   2.6    .8   2.1  1.0

2.91   Jason Terry     Dal   683   .615   22.7   2.3   4.1   2.2   1.3   1.6   .2
2.89   Jason Kidd      Dal   742   .570   11.0   5.4   9.3   1.8   2.0   2.8   .5
2.45 Russell Westbrook Okl   637   .490   23.3   5.5   7.5   3.2   1.4   4.5   .3
2.34   Chris Bosh      Mia   834   .556   19.1   8.9   1.3   2.1    .7   1.8   .8
2.11   Shawn Marion    Dal   691   .506   13.4   8.1   3.1   2.8   1.1   1.9  1.0

1.99   Zach Randolph   Mem   514   .523   20.7  10.5   2.5   3.2   1.0   1.9   .7
1.71   Tyson Chandler  Dal   681   .613   10.3  11.9    .5   4.2    .7    .8  1.0
1.65   Dwight Howard   Orl   258   .648   33.2  14.9    .4   3.6    .6   4.7  1.6
1.56   James Harden    Okl   534   .620   16.5   6.3   4.5   3.1   1.4   1.8   .9
1.56   Jose Barea      Dal   390   .514   17.4   3.9   8.1   2.5    .5   2.1   .0

Then just summing eWins from regular season and postseason + title bonus:

Code: Select all

e2W    equivalent wins  tm    RS     PO+       e2W  equivalent wins   tm    RS     PO+
21.7   Lebron James    Mia   17.7   4.06      12.7 Lamarcus Aldridge Por   12.0    .65
18.5   Dwyane Wade     Mia   14.3   4.12      12.1   Carmelo Anthony NYK   11.5    .67
18.4   Derrick Rose    Chi   15.4   3.04      12.1   Chris Bosh      Mia    9.8   2.34
17.9   Dirk Nowitzki   Dal   11.8   6.14      11.7   Paul Pierce     Bos   10.7   1.01
17.2   Dwight Howard   Orl   15.6   1.65      11.6  Amare Stoudemire NYK   11.4    .21
16.3   Kevin Durant    Okl   13.4   2.94      10.9   Al Horford      Atl    9.9    .98
15.5  Russel Westbrook Okl   13.1   2.45      10.8   Josh Smith      Atl    9.5   1.26
15.1   Kobe Bryant     LAL   14.0   1.15      10.7   Manu Ginobili   SAS    9.9    .72
13.9   Zach Randolph   Mem   11.9   1.99      10.7   Kevin Garnett   Bos    9.6   1.08
13.8   Pau Gasol       LAL   12.8    .90      10.7   Al Jefferson    Uta   10.7    .00
13.5   Blake Griffin   LAC   13.5    .00      10.2   Tony Parker     SAS    9.6    .63
13.4   Chris Paul      NOH   12.1   1.21      10.0   Luol Deng       Chi    8.5   1.50

Dirk missed 9 games. If he plays 77, he finishes 2nd.

Posted: **Sat Jun 25, 2011 6:18 am**

With the help of the additional playoff EWins, 6 of the top 7 players were on conference finalists but only 2 more players on a conference finalist are in the top 24 for a total of 8 and an average of 2 per team.

The losers in the 2nd round had 6 total and an average of 1.5.

A bit short of 60% of the top "players of the year" on the top 8 teams.

The losers in the 1st round had 7 total and an average of almost 0.9 per team.

The other 16 teams had 3 total and an average of less than 0.2.

Posted: **Sat Jun 25, 2011 12:28 pm**

Hey Mike -

I've been looking around for information on eWins, and I can't really find any. Do you have your method available anywhere?

Your metric seems to like D-wade more than others. I think I'm on board with that.

Posted: **Sat Jun 25, 2011 5:17 pm**

There are several recovered threads on eWins and several ways of finding them, including the index of topics thread. http://sonicscentral.com/apbrmetrics/vi ... ?f=2&t=250

Posted: **Sun Jun 26, 2011 7:29 pm**

Crow,

I've tried navigating that page, and honestly have no idea how to find anything really.

Posted: **Sun Jun 26, 2011 7:50 pm**

Tried to make it easier. Presented several different ways to search with instructions.

The index of topics thread has 10 threads that involve eWins that I could find immediately in the index by putting "eWin" in the 'find on this page' bar of my browser and clicking down the results. Several of the threads provide at least partial descriptions of the eWin formula and components and other background on how the data is processed to an eWin rating. I indicated where to look for those threads as best I could under the current constraints of the site. Seems easy enough to me but I put it together. I can't think of anything to make the method any clearer though. I'm sort of doubting I'll be updating the index without indications of frequent successful use.

On your specific interest, perhaps Mike G will re-explain it all again.

Posted: **Mon Jun 27, 2011 12:01 pm**

bbstats wrote: I've been looking around for information on eWins, and I can't really find any. Do you have your method available anywhere?
.

The theory: Boxscore statistical productions create a quantifiable potential for winning, and this is constant from team to team.
These stats are standardized: team and opponent production rates are scaled to standards. (I use 100 pts and 44 reb per team per game.)
Then a 'replacement player' productivity level is subtracted. The remainder is what wins games.

The method: Standardized per minute rates are weighted and summed to arrive at a total productivity: T
eW = (T - Tr)*Minutes/X
X is a constant across the sample (league, playoff series)
Tr is the replacement T. It's very close to 1/2 of league avg.

There are many adjustments to various rates (Sco, Reb, etc). I show them as per-36-minute, but they include: an estimate of the % of minutes vs starters, estimated % of points unassisted, estimated homer assist and block inflation, the % of assists on 3-pointers, ratio of shooting% to a standard (.527), opponent PPG, team Reb%, and etc.

This hardly suffices as an explanation of method, but that's a bit beyond me in finite time.
Even eMailing the spreadsheet to serious APBRMetricians, no one has managed to understand the method of eWins.

A team's players' eWins are added, and the team's expected wins are then given by:
xW = eW*2 - G/2
A number of parameters are adjusted so that xW = Pythagorean expected wins (pW), or as close as possible for teams in the sample.
In a playoff series, an exact fit is achieved: A steal, a rebound, etc, has the same value to both teams.
In a whole league, there are some differences. This last season, team xW erred (differed from pW) by an average of 1.30 in 82 games.

The formula above says that typically, a 61-win team has 51 eWins, and a 21-win team has 31 eWins.
The 61-21 team doesn't have almost 3 times the talent on their roster, but about 2/3 more.
eWins effectively splits the difference between point differential and win differential.
A player's annual fluctuation in eW/Min is a bit over half that of WS/Min, on average.

Now you must ask questions for further clarification.

Posted: **Tue Jun 28, 2011 12:14 pm**

bbstats wrote:...
Your metric seems to like D-wade more than others. ..

Maybe so. I don't know of many very similar metrics.
Here are a few years of eWins and Win Shares:

Code: Select all

year          ew/48   ws/48
2008    RS    .153    .082
2009    RS    .285    .232
2009    PO    .216    .195
2010    RS    .260    .224
2010    PO    .302    .229
2011    RS    .246    .218
2011    PO    .240    .216

In '08, Wade missed 31 games, and the Heat went 15-67, led by Ricky Davis.

Posted: **Mon Jul 11, 2011 6:47 am**

Ewins is respectable but how would you rate its ability to parse defensive greatness? I think that's always a big problem with statistics, James seems a little underrated there.

Tyson Chandler was second in Dallas playoff win shares, but he has less ewins than Terry, Kidd, and Shawn Marion.

I know a lot of guys here loved Varejao back in Cleveland, but I never bought into the notion he was their best defender. APM validates me, along with other figures.

APBRmetrics

2011 Player of the Year?

2011 Player of the Year?

Re: 2011 Player of the Year?

Re: 2011 Player of the Year?

Re: 2011 Player of the Year?

Re: 2011 Player of the Year?

Re: 2011 Player of the Year?

Re: 2011 Player of the Year?

Re: 2011 Player of the Year?

Re: 2011 Player of the Year?