Wins Developed: Projecting Player Improvement with Real-Plus-Minus


Oct 18, 2014; Atlanta, GA, USA; Detroit Pistons center Andre Drummond (0) blocks the shot of Atlanta Hawks guard Jeff Teague (0) in the fourth quarter at Philips Arena. The Pistons defeated the Hawks 104-100. Mandatory Credit: Brett Davis-USA TODAY Sports

I would like to apologize in advance for the belated nature of this post. Better late than never!

During the offseason, I reworked my Real-Plus-Minus (RPM) projection system, which I used in turn to give me Projected Team Win Totals. The crux of the system — the most detailed work I put into these projections — was predicting 2014-15 player RPM. A brief rehash of the method, in which I tried to find the accurate and statistically significant predictors of future RPM:

  • Prior 2 seasons of RPM, with the most recent season weighted more
  • Age (using a curve similar to this one)
  • Three extra-significant box-score statistics: blocks, steals, and missed field goal attempts. The latter of which being a negative of course.

Using this information I came up with what looks, to my eyes, like decent projections for every NBA player. Rather than post the entire list here, I wanted to show off the players who I project to do significantly better this season and significantly worse. Here’s the main tendencies I think you’ll see in RPM change, based on the three things mentioned above:

  • Younger players progress, older players regress
  • Shooters tend to regress significantly
  • Bigs tend to progress or stay consistent the most
  • Point guards tend to progress greatly and be loved by RPM
  • Player-seasons with a HUGE improvement (or decline) are regarded somewhat as a fluke

The best way I found to measure this is through what I call “Wins Developed”. According to my data (and conventional wisdom), player contributions are going to slide towards average (zero-RPM) every season. By subtracting this typical regression from my model’s projected RPM, and multiplying by how many minutes I project the player to be on the court, I estimate “Wins Developed.” This measures how many more wins a player will contribute to their team than a player that does not improve or worsen.

So here are the top 25:

Thoughts: As you can see, the Jazz have four of their biggest pieces looking like definite improvements, though Hayward still just looks like a small-market max contract guy rather than a large-market max contract guy. Kanter is still below-replacement level. Andre Drummond is definitely leading the “young bigs” explosion predictions, as his 2014 RPM figure was probably a little conservative. Ben McLemore is the worst player on this list, but as a 20-year old he figures to regress upwards towards zero more heavily. Millsap’s figure is perhaps the most intriguing to me on this list as he is past his prime. He definitely regressed last season in a “fluky” manner, so I see him gaining some more traction.

and bottom 25:

Thoughts: It is worth noting that I used different data and regression for Derrick Rose and Kobe Bryant as they missed significant minutes the last two seasons, so you might consider this my system being “one win dumber” if you’re fans of them. Great older players that don’t really put up huge box score numbers figure to regress hard: The Collison/Bonner/Frye/Tolliver/Iguodala are notables. Chris Paul’s age and style look to regress his 2nd-best in the league figure — and the same can basically be said of LeBron (who I still project to lead the league in RPM). LeBron is also losing some of his value from my projected diminishing returns for Cleveland’s offense [ED Note: If the idea diminishing returns in Cleveland reducing LeBron’s value seems counterintuitive, check out Nathan and I talking it through in the latest episode of The r-squared Podcast]. Perhaps most frightening on this list is Hardaway Jr and Mullens’ inability to improve more than typical player’s Horrible-RPM-Season.

Send me your anger/joy/questions @bbstats!