Pages Navigation Menu

Pacers and Wins Produced

Recently, Dave Berri, at the Wages of Wins Journal, put out a call looking for bloggers to use his statistical model, Wins Produced, to cover each NBA team. I volunteered to cover the Indiana Pacers for Berri’s “Wages of Wins Network.” My first post went up last week, and I thought it might require more explanation than was appropriate for that forum.

I have been reading Berri’s blog for over a year now. I am in the process of re-reading his book, The Wages of Wins, and look forward to immersing myself in his second book, Stumbling on Wins. His work is controversial and has caused a great deal of discussion across the internet. This FreeDarko post, and the accompanying commentary, is one of the most comprehensive discussions of his work I could find on the internet (although it falls mostly on the negative side). Like many basketball fans, many of his conclusions don’t fit with other pieces of my basketball knowledge. Some of what he writes makes complete sense to me, and other ideas challenge my conceptions. Over the past year I have come to a comfortable accord with his work: I see it as a valid statistical tool, one of many that I would like to keep in my analytical tool box. Like any other statistic, Wins Produced has strengths and flaws, answers some questions and raises others. It is one lens with which to examine the nature of basketball in an attempt to understand it better. In part, I volunteered to write with his blog as a chance to work with his statistical model, and hopefully understand it better.

After writing my first Pacers post for the Wages of Wins, I wanted to include some additional information. It seems that most of his blog readers have a working knowledge and a general acceptance of his ideas, methods and conclusions. I know that there are many others, Pacers fans included, who are much more skeptical. I don’t have the mathematical skill to defend or criticize his methods, but I would like to talk about some of the things that make sense to me when using the Wins Produced model to examine the Pacers. Again, I don’t advocate Wins Produced as a perfect or singular statistical model. But looking at the Pacers, some pieces of Berri’s model should make sense to even the most skeptical readers.

Let’s start with some basics about Wins Produced. Wins Produced and its more commonly used sister stat, Wins Produced per 48 (WP48), are formulas which use a player’s 2 Point Field Goals Made, 3 Point Field Goals Made, Free Throws Made, Offensive Rebounds, Defensive Rebounds, Missed Free Throws, Missed Field Goals, Turnovers, Steals, Assists, Blocks and Personal Fouls, to calculate how much they contribute to their team’s wins. How Wins Produced is calculated is less important to this discussion here than the statistical categories which are used, but if you are more curious here is a lengthy description of the calculation process

The impetus for the formation of this new “Wages of Wins Network” was the work of a blog reader Andres Alvarez, who created an automation for Berri’s Wins Produced. Previously each calculation was done by hand, by Berri for each post, meaning that league and team wide data was available only when Berri chose to cover that team for a post. The sticking point was a numeric adjustment for the position a player spent minutes at. This had been done by Berri’s subjective estimation, but Alvarez created a formula using a player’s height, weight, body-mass index, listed position and assist rate to automate a player’s position designation. While this formula is not perfect, it did allow for a database to be created showing Wins Produced and WP48 for each player in the league over the past few seasons. The position designation is important because the each player’s raw WP48 needs to be adjusted to account for the expected production from their position. To quote Berri:

As noted in The Wages of Wins, centers and power forwards get rebounds and tend not to commit turnovers.  Guards are the opposite.  The nature of basketball is that teams need guards, small forward, and big men.  Given nature of the game, players have to be compared to their position averages.

When calculating WP48 using the position adjustments, every minut a team played needs to be accounted for by all five positions. The Pacers played 19,705 minutes this season. Which means 3,941 minutes were available for each position. If you consider that some of Danny Granger’s minutes were spent at power forward, then you need to account for who was playing small forward during those minutes. When looking at Alvarez’s site you will see the position designations as assigned by his formula expressed as a percentage. For example his formula has Mike Dunleavy spending 52% of his minutes at shooting guard and 47% of his minutes at small forward.

When look at WP48 there are some general benchmarks to keep in mind. Anyone with a negative WP48 would be considered to have hurt his team more than they helped. Essentially, their play was producing losses instead of wins. A WP48 of 0.100 is considered average, anything beyond that is obviously above average. A WP48 of 0.200 or higher is considered “star production.” A WP48 of 0.300 or higher is considered “super-star production.”

These benchmarks create much of the skepticism from Pacers fans. In his time in Indiana, Troy Murphy has consistently posted WP48 numbers above 0.200, while Danny Granger has hovered around 0.100. “There is now way Murphy is our best player, let alone a star. There is no way Granger is merely average,” are common refrains. What I want to do is look at some numbers and see where these WP48 numbers come from when looking at the Pacers.

In the analysis below I compared each player on the Pacers roster to an average player at their position, using the statistics which are components of Wins Produced and WP48. For the sake of simplicity I have just used a player’s listed position for comparison and calculation. This explains why some of the WP48 numbers here won’t match with the ones on The Wages of Wins or Andres Alvarez’s Wins Produced Automation. My hope is that by looking at the WP48 numbers for the Pacers not just in isolation, but as a reflection of how they compare to average players in the component categories, they will make a little more sense and connect more closely to our subjective observations.

*Below Average Numbers in Red
*Numbers are per 48 minutes
*Points-per-shot = [PTS-FTM]/FGA
*Adjusted Field Goal Percentage = PPS/2
*Net Possessions = Rebounds + Steals – Turnovers
*Win Score = PTS + REB + STL + ½*BLK + ½*AST – FGA – ½*FTA – TO – ½*PF

  • Roy Hibbert is classified as a below-average center by Wins Produced, and looking at these statistical categories, it’s easy to see why. Hibbert is significantly below-average in the possession categories, specifically rebounding, steals and turnovers. Despite a dramatic improvement last season he is still below-average in Personal Fouls per 48 as well.
  • Jeff Foster was injured most of last season, but performed reasonably well in his limited minutes. His terrific rebounding ability helped to offset some of his below-average production in the scoring categories.
  • According to WP48, Solomon Jones was the worst player on the Pacers roster last season, and here it is easy to see why. He offers significantly below-average production in almost every statistical category used to calculate Wins Produced. 


  • When using only the power forward position adjustment for Troy Murphy he comes out at 0.321, which classifies him as a super-star by Wins Produced. This is confusing for many people, but when looking at these statistical categories it makes more sense.  His 3PT shooting ability makes him significantly above-average in terms of Points Per Shot and Adjusted Field Goal Percentage. He is also significantly above-average as a rebounder. In fact the only categories where Murphy’s production is below-average for a power forward is in the areas of Free Throw Attempts and Blocked Shots. When you factor in a position adjustment for the minutes Murphy has played at center his WP48 comes down a little bit, below 0.300.
  • Tyler Hansbrough had a surprisingly good rookie season, when he was able to make it onto the floor. He played with a frenetic energy which allowed him to draw a lot of fouls and do an unexpectedly good job of rebounding. Unfortunately, this frenetic energy made it difficult for him to finish effectively and he shot only 36.0% from the field. This number is responsible for his extremely low Points Per Shot and Adjusted Field Goal Percentage numbers, and in turn his very low WP48.
  • Josh McRoberts was very productive especially towards the end of the season when he finally started receiving consistent minutes. McRoberts was below average in a few categories, but prove to be a very efficient scorer, and above-average in essentially all the possession categories. He has definitely proven he deserves more minutes next season.

  • Danny Granger is considerably above average in all of the scoring categories. He is below average with respect to Rebounds, Turnovers and Win Score. His relatively modest performance in these possession categories is why Wins Produced views him as a simply above-average performer as opposed to the top-tier star many fans believe him to be.  When you factor in the minutes he plays at power forward in small lineups, his WP48 decreases a little more as his performance in the possession categories is even further from average.
  • Mike Dunleavy was just slightly below-average in a few categories. When you factor in the minutes he played at shooting guard his WP48 increases slightly to an essentially average player.

  • When Brandon Rush is compared to the average shooting guard his statistics look very reasonable. He is below average with regards to Free Throw Percentage, Field Goal Attempts, Free Throw Attempts, Points Scored, Steals and Assists. He is above average with regards to Points per Shot, Adj. Field Goal Percentage, Rebounds, Turnovers, Net Possessions, Blocked Shots, Personal Fouls and Wins Score. All of this balances out to an average performer at the shooting guard position. Considering Brandon Rush to be an average shooting guard is a leap for many Pacers fans, but when focusing on these statistical categories it seems more reasonable.
  • Dahntay Jones’ performance is strikingly below-average, almost across the board.
  • Luther Head is above average in most of the shooting/scoring categories, but is below-average with regards to most of the possession categories. 


  • Wins Produced viewed Earl Watson as the Pacers best point guard last season, and looking at these statistics it’s easy to see why. While he didn’t score as much as Ford or Price, he was above-average with respect to Rebounds, Steals, as well as being the Pacers best assist man.
  • Price was above-average with respect to the scoring categories, and very close to average in most others. His paltry 5.9 Assists per 48 minutes almost completely explains his below average showing in WP48.
  • I think most Pacers fans would disagree with the idea the T.J. Ford was essentially just as productive as A.J. Price this season. But again . . . look at the numbers here. While Price was above-average in the scoring categories, Ford proved to be above-average when looking at Rebounds, Points Scored, and Free Throw Attempts. While still below-average, Ford was also much more effective distributing the ball than Price.

To avoid confusion, let me restate the argument here. I am not saying that Wins Produced is the best or only statisical model with which to view basketball performance. I would like to point out that, with respect to the Pacers, the numbers make a lot more sense than we usually give them credit for. When you break them out into their component categories and compare each players production to an average NBA player, it is easier to see why Troy Murphy looks like a star, Danny Granger looks closer to average, and Roy Hibbert looks woefully underproductive. Whether these statisical categories accurately or completely reflect a player’s true worth is a discussion for another time and place.

I would encourage Pacers fans with even a passing interest in statistics to read The Wages of Wins and sample a few posts at the accompanying blog. Ask some questions in the commments there and at other basketball stats sites you enjoy, dig a little deeper, and look at some more of the numbers before you decide whether Wins Produced can help you to a better understanding of the NBA.

  • Pingback: Not So Final Thoughts on the 2010 Finals « The Wages of Wins Journal

  • robbieomalley

    Hey man, how did you calculate PPS and Adj FG%? For some reason when I calculated it for Ford and Watson, I got different numbers.

    • ilevy

      I went back and double checked to make sure. I took my stats from Basketball-Reference. Unless I am mistaken the formula for Points per Shot is Points Scored – Free Throws Made/Field Goals Attempted.

      For Ford – 484 (Pts) – 104 (FTM)/ 418 (FGA)= 0.9090, which I rounded off to 0.91 here. Divide that by 2 for Adj FG% and you get 45.45, which I rounded to 45.5%.

      For Watson – 619 (Pts) – 120 (FTM)/ 524 (FGA)= 0.9522, which I rounded off to 0.95 here. Divide that by 2 for Adj FG% and you get 47.61, which I rounded off to 47.6%

      Maybe I misunderstood the formulas, but I think I’ve got that right. If you see a mistake I’ve made, please let me know. I am still pretty new working with some of these numbers. Thanks for reading and commenting!

  • Xavier Q

    Lifelong Pacers fan and a big fan of Prof. Berri. The deviation between what WP48 shows and what fans “know” is because of the emphasis on scoring totals. In the NBA you judge a player first by his points per game. WP works under the premise that the net possessions you get for your team is the most important factor, followed by shooting efficiency, not points scored. For example, if Danny shoots 42% and takes 20 shots, yet Roy shoots 50% and takes 12 shots, it’s reasonable to say that some of those shots should have been shifted over to Hibbert.

    I’m not saying that Prof. Berri is absolutely right with his model, but given that WP accurately models how many wins a team will get in a year, it’s a compelling argument.

    • ilevy

      Thanks for reading and commenting! I agree with almost everything that you said, and appreciate such a succinct summary of the argument for Wins Produced that I sort of danced around in my post.

      One of the few things I struggle with when looking at Wins Produced, and the idea you brought up with Hibbert and Granger is the idea of static production. From what I have read by Dr. Berri, aside from some increases due to experience at the beginning of a player’s career and some age-related decline at the end of player’s career, most production is static from year to year. I agree with this assertion for the most part, but I struggle with the idea of a player’s production being consistent when they recieve increased minute or opportunities.

      I have seen several Wins Produced related posts where a player’s numbers in limited minutes are used to advocate for them receiving more playing time. “Player A is a back up but has a higher WP48 than the starter, so maybe he should become the starter.” I think this really undervalues the amount that role and situation factors into production.

      Looking again at the case of Granger and Hibbert. Hibbert has a higher FG% than Granger. In the present that means he has been a more efficient scorer than Granger, but I don’t know that indicates he would be a more effective scorer than Granger in increased opportunities. I believe that Hibbert’s FG% is largely a function of his shot selection and willingness to pass out of situations where he doesn’t have a good shot. Asking him to take more shots ignores the issue of whether he and the team actually have the ability to create more shots for him of the same quality as the ones he already takes. It seems very possible that asking him to take more shots, would lead to him taking some lower quality shots, decreasing his FG% and defeating the purpose (Offensive Efficiency) of having him shoot more in the first place.

      I am not saying Hibbert shouldn’t take more shots and Granger shouldn’t take less, in fact I have been quite critical of Granger’s shot selection here and at Indycornrows as well. I am just very hesitant to use WP48 numbers to make blanket projections about who should receive more or less playing time or shot attempts. I think it is incumbent on us as fans to understand what situations have allowed players to produce those numbers and include that understanding in our suggestions and recommendations for roster changes.

      Thanks again for reading and commenting, glad you enjoyed the post!

      • tgt

        As you noted, there has been much discussion about how efficiency changes with usage. I believe it was B-R that researched this and came up with the general rule that each 1% more a player is used offensively, they will be 1% less effective. If someone normally uses 15% of the possessions when they’re on the court, and scores 1.2 points per possession, then if they use 20% of the possessions, they are likely to score around 1.15 points per possession. I’m not sure I’m sold on the accuracy of their methodology, but it’s a starting point. Also, note that the stat points per possession is different from points per shot. Possessions include turnovers and free throws. Offensive rebounds factor in as well.

        The Rockets and Grizzlies were great test cases of putting together lots of low usage players and lots of high usage players and seeing how their efficiency changed. I don’t remember any follow-ups by Berri or elsewhere about how this turned out, and I haven’t put in the time to check it myself.

        Returning to the point of Granger vs Roy, while I would not say all of Granger’s attempts should be moved elsewhere, attempting to move 5% or 10% of his attempts would likely be beneficial. This could be the result of running 2 or 3 plays a game through someone else. Assuming they defer 1 to 2 times due to bad shots, the overall rates shouldn’t change that much.

%d bloggers like this: