Tuesday, June 10, 2008

FAQ about statistics used on Bucks Diary


NOTE 1: Most every statistic on Bucks Diary is somehow based upon the Win Score metric invented by Professor David Berri based upon the research he did for his book, The Wages of Wins. His work is to basketball as Bill James work is to baseball and the Football Outsiders work is to football. Win Score, I believe, it is THE one performance metric that most closely explains, through statistics, how a player performed.

NOTE 2: Win Score is a compilation of statistics, weighted according to how each statistic correlates with victory. The Win Score weighting system has a 96% correlation with wins. In other words its pretty accurate.

3. What is the Win Score Formula?
Points + Rebounds + Steals + 1/2 Assists + 1/2 Blocks - Turnovers - Field Goals Attempted - 1/2 Free Throws Attempted - 1/2 Personal Fouls = The Player's Win Score Points

4. How do you find "Win Score Above Average"?
You take the player's Win Score and divide it by the minutes he played, and then multiply that by 48. Then you adjust that number according by the average Win Score compiled at the position the played. If he played multiple positions, you have to adjust your average accordingly.

5. Can you do an example?
Sure. Let's say I look at the box score and figure out Kobe Bryant's Win Score added up to 8.5 Win Score points. Then I look at his minutes played. It was 32. Okay, so that comes out to .265 Win Score Points per minute, for a Win Score per 48 minutes of 12.3. Now I must adjust for position. Usually I do this by using the player's positions as recorded on 82games.com. According to 82games.com, Bryant plays 75% of the time at SG, and 25% at SF. The Win Score positional average for SG is 6.4 and SF is 7.8, which works out to an average of 6.8 for the posititions Kobe played. So, for the game Kobe had a Win Score Above Average of +5.5.

6. What is Opposition Win Score?
Now this one is my invention, a defensive metric I conjured based on Win Score, but not yet quite endorsed by Professor Berri, but I think its supported by math and logic. What I do is I take the "Opponent Counterpart Statistics" on 82games.com, which are the average statistics compiled by the players the particular player has guarded, and figure out the "Win Score Above Average" in reverse, essentially. I've found that if you take any given team's cumlulative Win Score and its Oppositition's Cumulative Win Score, adjust those for the NBA cumulative average, add the two numbers together, and then divide by 5, and then punch the result into the Win Score Win Prediction Formula, you will be able to predict the team's wins with 93% accuracy. So I think its a legitimate statistic.

7. What is Win Contibution?
This is my own invention based on Win Score and on baseball statistics invented by Baseball Prospectus, Hardball Times, Bill James, and others. As I have said, I believe it is the very best statistic available for evaluating a player's performance in any single game. Its meant to be the basketball equivalent of the baseball statistic "Replacement Value over the Average Player". Essentially all it does is it takes the player's "Win Score Above Average" and multiplies it by the player's percentage of total playing time. Thus, if I'm doing Win Contribution for a regulation NBA game, the formula : Player's Win Score Above Average * (player's minutes/240). Therefore, the average player (meaning a player whose Win Score per 48 is right at his positional average) would have a +0.000 Win Contribution. A team of player's who record +0.000 would be expected to have as many wins as they have losses. Thus any "Contribution" over +0.000 is a contribution to a winning team.

8. What is Offensive and Defensive Win Contribution?
This I use only when evaluating player's over a course of games. If I'm doing it that way, I'll seperate it into offensive and defensive Win Contributions. Offensive Win Contribution is the Win Contribution described in #7. Defensive Win Contribution is simply the negative value of the player's opponents Win Contribution. In other words, the formula is - Opposition Win Score Above Average * (the player's minutes/total playing minutes available over the relevant time period). To get to what I refer to as "Total Win Contribution", you add offensive and defensive Win Contributions and divide by 2.

9. How do we know which Win Contribution you are using?
Unless I indicate "Offensive" and "Defensive" Win Contribution, I'm using the straight Win Contribution described in #7. I always use #7 Win Contribution when evaluating individual games, because I will always provide both team's Win Contribution. That way, you can look at each and sort of tell who was guarding whom and do your "defensive" Win Contributions mentally. The basic reason for the two usages is over the long haul, I have access to 82games.com's "Counterpart" statistics, but they don't provide them for single games.

10. How do you translate Win Score or Win Contribution into wins?
Here's the formula: Win Score Above Average/48 * 1.621 + 0.120/48 * minutes played = wins. To translate "Wins Contribution" numbers into wins, you do the following, you add up the team's offensive and defensive Win Contributions, add those numbers together, divide that by two, and then you treat that result as if it were "Win Score Above Average" and punch it into the formula above.

11. What's a good Win Contribution Score?
If you notice, over the course of a season, most of the player's Win Contributions regress to the mean, which is +0.000, which is what you would expect. For one game, though, you can read Win Contribution as such:

below -1.000: awful game, can be blamed if we lose

below -0.500: very bad game

below -0.300: bad game

-0.100 to -0.300: below average game

-0.100 to +0.200: average game

0.200 to 0.400: good game

0.400 to 1.000: outstanding game

0.1000 and above: OUT OF HIS MIND, if we win, he gets the game ball

No comments: