This week I'll look at Dr. Dave Berri's Wins Produced stat, one of the more controversial and intriguing measures out there…and one of my personal favorites. Dr. Berri, a professor in economics, did a series of linear regressions (read: complex math) on several seasons worth of stats and came up with a statistical model that he says can explain 95|PERCENT| of a team's wins based upon the box score numbers of it's individual players. And as an added treat, Dr. Berri was kind enough to e-mail me with responses to the critiques of his methods that I mention (see Weaknesses, below) and even point out a couple of other critiques that may be more valid.
What is it:
At its simplest, "wins produced" is a measure that looks at offensive and defensive efficiency in one stat. The math is complex, and for those interested in the details you should check here. Briefly, offensive efficiency is defined as points scored divided by possessions used and defensive efficiency is defined as points surrendered divided by possessions used. These efficiencies, as well as several other factors, are combined to produce a player's calculated production relative to their position average. An "average" player has a "wins produced per 48 minutes" (WP48) mark of 0.10, and a "Win Produced" is defined as 0.10 + production relative to position average. Roughly speaking, a good player/starter has a WP48 higher than 0.20, a great player a WP48 over 0.30, and any WP48 approaching 0.40 or higher is historically good. At the other extreme, for very inefficient players it is possible to produce a negative number of wins (or play losing basketball).
Strengths:
The coolest aspect of "wins produced" is the concept of assigning a fixed amount of a win to box score statistics, similar to the Win Shares stat we discussed last week. While the math may be complicated, there is nothing easier to understand than a stat that says Chris Paul was responsible for 25.4 wins in 07-08, while Deron Williams was responsible for 15.2 wins. The reader does not have to guess just how much one player is better than another like in other one-number stats. Paul had a PER of 28.3 and Williams a PER of 20.8 last season…just how much more valuable is that according to PER, in practical terms? We see that with Wins Produced, it was exactly 10.2 games more valuable.
The flip side of the coin is that since the "wins produced" stat theoretically measures player value independently of teammate caliber, it can be used as a predictor for how a player and team should be expected to produce after player movement (i.e. free agency or trade). The most famous success story for Berri in this capacity is the Allen Iverson/Andre' Miller trade of 2006, which I have written about in the past. That trade was supposed to benefit
The concept of scoring being completely overrated as a component of winning basketball is also buzzworthy, because to most people scoring is the most important aspect of basketball. One of the "truisms" that I often hear is that at the end of the day, basketball is about putting the ball into the hole. The scorers tend to get the big contracts and the commercial deals, while players that specialize at areas outside of scoring are considered "role players". If it gains acceptance, though, "wins produced" would turn that concept on its ear. Any statistic that runs so counter to the prevailing notion, but stands up well over time as a good predictor of success has the potential to completely change the way we view, analyze, and teach basketball…this would be a pretty big strength.
Finally, because the statistical model was developed from mathematical analysis of actual wins and losses, there is explainable logic for the formula. This is different from PER, for which John Hollinger came up with the weights and values for each stat category based on his own intuition and thus isn't as mathematically sound. This also separates Wins Produced from Win Shares, or other similar metrics, and helps it overcome scrutiny when the measure yields a non-intuitive conclusion.
Weaknesses:
1. One of the more common criticisms of "wins produced" is that many feel that it overvalues rebounding. In Berri's system, the player that gets a defensive rebound is given credit for ending an opponent's possession and re-gaining possession for their team. But that could be giving too much credit to the rebounder and not enough to the on-ball defender that may have caused the missed shot in the first place. As such, big rebounders like Marcus Camby are always near the top of the rankings. One result of the premium put on rebounding is that, according to Wins Produced, Dennis Rodman was the most valuable player on the 1998 Bulls instead of Michael Jordan. Needless to say, for many this is too radical a concept to pass the sniff test and has earned Berri some criticism.
Dave Berri Response:"In November of 2006 (over two years ago) I posted the following: Do We Overvalue Rebounds? What this post notes is that even if you lower the value of a rebound by 30|PERCENT|, the story we tell about players doesn't really change. We can go ever further than this argument.
According to Wins Produced and Position Adjusted Win Score, Ben Wallace was the most productive player in 2001-02. If you reduce the value of a rebound from 1 to 0.5, Ben Wallace is still the most productive player that season. So very large changes in the value of rebounds will not change the basic story being told.
I would add that the one factor that has the largest impact on WP48 is shooting efficiency. And that is the key insight of Wins Produced. Efficient scoring is very important to a team's success in the NBA."
2. Another criticism for the "wins produced" stat is that Berri employs a "team adjustment" factor to his numbers, which some believe is a fudge factor that allows Berri to massage his calculations to fit different circumstances. The only real way for "wins produced" to address criticisms like this is by tracking performance over a long period of time…if it is proven correct as a predictor enough times then questions about a fudge factor would become moot, whereas if it repeatedly fails and "fudges" then eventually it could hurt the credibility of the stat.
Berri response: "The team adjustment is not a "fudge" factor. Offensive and defensive efficiency are comprised of statistics for the player and the team. The team factors are associated with defense. The underlying assumption in calculating Wins Produced is that defense is a team activity. Therefore it makes sense to allocate the team defensive factors across players according to minutes played. This means that players on different teams get credit for the defense played by their respective teams.
One could also use plus-minus data to allocate the team defensive values. I experimented with this in the past, but more work can be done. Regardless of the approach to team defense, it is not a "fudge" factor. In other words, it is not something I threw into the model to make it explain wins."
Dr. Berri's suggestions for possible weaknesses
1. Wins Produced does not give credit to players for creating shots. This is generally perceived as the biggest weakness. This perception is based on two beliefs. a) it is difficult to get a shot off in the NBA and b) the more you shoot the lower your efficiency. I am not sure there is much empirical evidence behind either proposition. Plus, if you give credit for taking shots then inefficient shooters will look better. Nevertheless if you believe shooting is difficult, Wins Produced will disappoint. And connected to this point…if you believe scoring is the most difficult thing a player does in the NBA, and therefore scorers are the most valuable, Wins Produced will disappoint also.
2. How vs. Why: Wins Produced tells you how productive a player is, but it does not tell you why. Performance is impacted by age, injury, roster turnover, coaching (in a few cases), and the productivity of teammates. The latter issue is the idea of diminishing returns. The more productive your teammates, the less productive you will be. This is a real effect in the NBA, although in the aggregate it is rather small. Still, these effects are not part of Wins Produced and these issues can impact what we see in the future from a particular player.
Usage:
Wins produced styles itself to be a total player measure, so in theory one could simply say "player X had more wins produced than player Y so he is better"…but practically, by now we should all know that there is no holy grail stat that is always right. You should still corroborate the wins produced output with a variety of other stats and personal observations before making a final conclusion. Keep the pros and cons of Wins Produced in mind, and decide in your own head how much weight to give to the results of this measure when making your own evaluation.
Quick Links for series:
Intro
PER
+/-
Offense/Defense Rating
Kobe vs. LeBron by the numbers
Win Shares