Plus/Minus Game 6: Spurs 103, Hawks 88

[FromWayDowntown]

Again, not much surprising about these numbers, which show that the Spurs' biggest surges came with the starters on the floor, while also showing that the Hawks were able to make up some ground on the Spurs' bench. Anyway, here are the plus/minus from last night:

+24 - Nesterovic
+21 - Ginobili
+19 - Parker
+18 - Bowen
+18 - Duncan
+3 - Horry
-1 - Barry
-3 - Brown
-4 - Udrih
-5 - Wilks
-6 - Massenburg
-9 - Rose

And here are the cumulative numbers through 6 games:

+69 - Ginobili
+65 - Duncan
+48 - Nesterovic
+44 - Bowen
+33 - Parker
+19 - Horry
+8 - Udrih
+4 - Wilks
-3 - Barry
-5 - Brown
-8 - Rose
-24 - Massenburg

My .02 on this: I think anyone using this statistic has to take it with a grain of salt. I certainly don't think that it necessarily accurately reflects the quality of a player's contributions. It's an effort to account for things that don't necessarily show up in a box score, but things that a savvy fan would sense or notice -- I think Horry is a good example of this: his gross statistics will never be particularly good, but when you look at him from a plus/minus perspective, Horry is almost always in the positive. So, in that sense, Horry's traditional statistics don't accurately reflect his impact. It's true that, taken in isolation, the number could just as easily mean that a player just happens to be on the floor when good things happen as it could mean that the player is responsible for those good things happening. But, it's a step towards analyzing things that are hard to measure. This statistic is also unfair to some of the bench players who have been asked to play some garbage time minutes, thus costing them some late slippage. But, over the course of the season, the plus/minus begins to give some sense of whether a player's contributions have generally been positive to the team's success.

I've been trying to work out plus/minus numbers for the various groups that have played together -- looking at it from the standpoint of groups and combinations is probably even more telling, and when I get enough time to cull those numbers, I'll post the results and try to keep that updated from time-to-time.

[deepsouth]

Dowtown.
How do you obtain the basic data for every game?
Do you collect it yourself? It is very interesting.

[Rick Von Braun]

FWD,

I am glad another fellow Spur fan is inclined to take a look the stats of the game.

The pure plus/minus statistic is useful, but not very accurate since it doesn't consider the quality of your own teammates and the quality of the opposition. In addition, it doesn't measure the quality of the player per se, but the relative difference of the players that substitute in for him.

Adjusted plus/minus stats have been the standard in basketball statistical analysis for some time. The adjusted stats consider the quality of the teammates and the opposition, the differences between playing in the road or home, and the differences between garbabe or clutch time.

Jeff Sagarin (http://www.usatoday.com/sports/sagarin/nba0405.htm) has applied these concepts in his WINVAL metric, measuring team peformance and individual peformance in non-team sports (such us golf, tennis, etc.) for several years. He is reportedly being paid more than 100K per year by the Mavericks to run the WINVAL system for them.

Going even further down the road, Dan Rosenbaum (http://www.uncg.edu/bae/people/rosenbaum/nba.html) expanded this metric with a hybrid system that combines the adjusted plus/minus with a statistical index derived from the pure adjusted plus/minus. He called this metric WINVAL2 (http://www.uncg.edu/bae/people/rosenbaum/NBA/winval2.htm). This metric is very good. It has very little noise and very few statistical outliers.

If you are really interested in taking a serious look at the stats of the game, I can help you with some pointers and advice. The last reference requires some knowledge of linear regression analysis to understand how the metric is derived. Don't get discourage if there is something you don't understand. Drop me a PM, and I'll be glad to help you.

Edit: fixed the URLs.

[xcoriate]

^^ Whoa....... that shit is complex

[FromWayDowntown]

RVB --

I'm down with the more complex methods for analyzing the data and drawing better conclusions from what the data show. Frankly, if I had more time, I'd tool around with more sophisticated calculations to try to more closely ascertain player value -- whether by using valuations similar to Sagarin's or Rosenbaum's. But, given my relative lack of sophistication with computers and the substantial amount of data that such such calculations require (meaning that I'd have to do most of the calculation/data compilation by hand), I'm pretty much stuck to keeping this on a pretty basic level.

I will read the articles you've recommended when I have an opportunity, though. Thanks much for pointing them out to me.

For now, I'm intrigued enough by the simple "hockey form" plus/minus and find that I can substantiate most of what I see with that statistic. If people don't find it useful (or would rather that I stop posting it because of its imperfections) I'll stop. For the purposes of a night-by-night analysis, that seems to be a somewhat useful measure and it seems that some around here are willing to overlook its acknowledged inadequacies in favor of another means of measuring performance.

Like I say, I'm going to put together some group plus/minus figures in the next few days and will post those when I get a chance -- those should be more illuminating on where the actual gains and losses come about.



Dowtown.
How do you obtain the basic data for every game?
Do you collect it yourself? It is very interesting.

deepsouth, I do my own calculations from the play-by-play of the game.

[deepsouth]

FWD:

Thanks for the answer.
I asked it because I am really interested in statistics (and in Spurs), and I thought if there wasnt possible to acquire the basic data from the yahoo game channel (or similar play by play web page).

[Rick Von Braun]

RVB --

I'm down with the more complex methods for analyzing the data and drawing better conclusions from what the data show. Frankly, if I had more time, I'd tool around with more sophisticated calculations to try to more closely ascertain player value -- whether by using valuations similar to Sagarin's or Rosenbaum's. But, given my relative lack of sophistication with computers and the substantial amount of data that such such calculations require (meaning that I'd have to do most of the calculation/data compilation by hand), I'm pretty much stuck to keeping this on a pretty basic level.Please don't spend a single minute trying to do this by hand. People use some stats package to crunch the numbers, like SAP or R (I personally use the latter). http://www.spurstalk.com/forums/images/smilies/smielephant.gif


I will read the articles you've recommended when I have an opportunity, though. Thanks much for pointing them out to me.I repeat my offer to explain in basic terms what the articles mean. If you are stuck, drop me a PM.


For now, I'm intrigued enough by the simple "hockey form" plus/minus and find that I can substantiate most of what I see with that statistic. If people don't find it useful (or would rather that I stop posting it because of its imperfections) I'll stop. For the purposes of a night-by-night analysis, that seems to be a somewhat useful measure and it seems that some around here are willing to overlook its acknowledged inadequacies in favor of another means of measuring performance.

Like I say, I'm going to put together some group plus/minus figures in the next few days and will post those when I get a chance -- those should be more illuminating on where the actual gains and losses come about.Please continue to do so... I found your stats very useful. They provide another point for comparison and discussion.

There are some stats available as well from 82games (http://www.82games.com/0405SAS.HTM).

For the Spurs, the On/Off Court pure +/- stats are:



Player On Court Off Court Team Net
+/- +/- +/-
Ginobili (http://www.82games.com/04SAS6D.HTM) +18.6 -9.0 +27.6
Duncan (http://www.82games.com/04SAS11D.HTM) +14.6 -9.8 +24.4
Nesterovic (http://www.82games.com/04SAS12D.HTM) +14.4 +0.4 +14.0
Bowen (http://www.82games.com/04SAS7D.HTM) +12.0 +2.9 +9.1
Horry (http://www.82games.com/04SAS9D.HTM) +15.4 +6.5 +8.9
Wilks (http://www.82games.com/04SAS3D.HTM) +12.5 +8.1 +4.4
Udrih (http://www.82games.com/04SAS1D.HTM) +6.2 +9.1 -2.9
Parker (http://www.82games.com/04SAS2D.HTM) +7.5 +10.6 -3.2
Brown (http://www.82games.com/04SAS5D.HTM) -3.7 +11.9 -15.6
Rose (http://www.82games.com/04SAS8D.HTM) -4.6 +14.6 -19.2
Barry (http://www.82games.com/04SAS4D.HTM) -0.9 +19.1 -20.0
Massenburg (http://www.82games.com/04SAS10D.HTM) -26.4 +14.6 -41.0

[Nikos]

Please don't spend a single minute trying to do this by hand. People use some stats package to crunch the numbers, like SAP or R (I personally use the latter).

What do you mean you use the latter "R"? Are you insinuating that you attempt to do your own DANVAL or WINVAL2 numbers? Those are very complicated unless you have access to advanced statistical programs and factor in the play by play of EVERY single NBA game (thru nba.com).

What type of analysis do you do RVB? Let me know, I am very interested.

[Rick Von Braun]

Due to my professional activities, I perform statistical anaysis all the time, so I am familiar with the tools.

Send me a PM and I'll fill you in. http://www.spurstalk.com/forums/images/smilies/smielephant.gif

[timvp]

Keep doing what you're doing, FWD. The +/- is a very telling stat.

[xcoriate]

Sure there are imperfections but as long as these are accounted for/understood by the reader its good and very informative. Keep it up please.