web analytics

Underrated and Overrated Via RAPM

November 2, 2012

In several of my recent articles, I have been using Jeremias Engelmann’s 12 year average RAPM dataset. Here, I will look at what players may have been underrated or overrated by even advanced stats over the course of the last 12 years.

I decided to use PER and WS/48 as my advanced stats to compare to, and I am limiting this to players with a very large sample size (though the user can adjust the threshold below). I’m using 20,000 possessions as the lower limit. At that level, RAPM should be very accurate in quantifying actual value to the team in points per 100 possessions.


What do we see? Both PER and WS/48 underrate a similar group of players: Bruce Bowen, Shane Battier, Luol Deng, Rasheed Wallace, Ben Wallace, Metta World Peace, Jared Jeffries, Thabo Sefolosha, Baron Davis (?), and of course Kevin Garnett. PER, as John Hollinger has said, does not attempt to quantify defense at all, so it makes sense that the underrated players for PER are the great defenders. WS/48 in general correlates better with RAPM (R^2 of 0.47 compared to 0.35), but similarly has difficulty quantifying defense. In general, no box score stat can get very close to accurately measuring defense.

WS/48 doesn’t do as badly with the defenders. It does pretty well with good defenders who mostly add value through blocks, steals, and rebounding, such as Chuck Hayes or Anderson Varejao.

In general, most of the underrated players by both PER and WS/48 are defensive big men (mostly) or the players you have heard of as great perimeter defenders (Deng, Kyle Lowry, Conley, Battier)


What about the overrated? Well, we know what PER overrates: scoring. Getting buckets. So we have great scorers with suspect defense, like Kevin Durant, Al Jefferson, Brook Lopez, Drew Gooden, Andrew Bynum, JaVale McGee, Carlos Boozer, even Karl Malone… Interestingly, most of the worst overrated offensive-minded big men. Glenn Robinson and Stephon Marbury are in there, also.

WS/48 overrates the same class of offensive big men, just not as badly. Carl Landry, Marcin Gortat, Troy Murphy, plus all of those mentioned by PER.

What’s the takeaway?

Box score stats miss defense. Defense matters more for big men; the variance for DRAPM seems to be greater for big men than for guards. In general, offensive/scoring big men are overrated and defensive big men are underrated. It seems that guards are rated pretty well, in general, though lock-down guards and swingmen are also underrated.

Kevin Garnett and Tim Duncan have been ridiculously good over the last 12 years. Put either of them on an otherwise average team and that team becomes a title contender, if not winner.

Most underrated:

Most overrated:

See Part 2 of this series at Overrated and Underrated via RAPM (Part 2)

Tags: , , , , ,

10 Responses to Underrated and Overrated Via RAPM

  1. [...] Daniel Myers happened to do a good post today [http://godismyjudgeok.com/DStats/2012/nba-stats/underrated-and-overrated-via-rapm/] comparing PER and Win Shares to Regularized Adjusted Plus-Minus to show how players are underrated [...]

  2. Ben on November 3, 2012 at 9:21 am

    Great lists – would be interesting to break down by offense and defense.

    Also, Durant really stands out as one that doesn’t fit my intuition. I remember his rookie +- was really terrible. I wonder what would happen if that season was just wiped from the record.

    • DanielM on November 3, 2012 at 10:23 am

      I could certainly break it down by offense and defense, though PER is not a measure of defense at all.

      Since this is “average”, everything is effectively weighted by minutes played. In his first season or two, KD was a bit of a chucker on offense–he has certainly improved his efficiency and defense a lot.

      • Ben on November 5, 2012 at 9:23 am

        I was thinking more of using OWS and DWS rather than WS and PER, or even offensive and defensive versions of SPM.

        I would guess that some of Durant’s improvement got captured by Westbrook and Harden in the adjusted plus-minus stats since it coincided with their arrival.

        • DanielM on November 5, 2012 at 12:54 pm

          Sounds like a good idea for a follow-up post.

  3. travis on November 4, 2012 at 5:44 pm

    i always thought great defense was underated. example, jared jeffries was a core member of their rotation. he had to back up stoudamire and chandler, and with all the knicks scores and shooters he dosen’t need the ball to make a difference. every team needs guys like these.

  4. Overrated and Underrated via RAPM (Part 2) | DStats on November 10, 2012 at 6:28 am

    [...] Underrated and Overrated Via RAPM [...]

  5. Daniel on February 12, 2013 at 10:03 pm

    I’m super late replying here, but have you considered looking into players rated by a crowd(e.g., the ESPN top 500 panel rankings) and looking for overrated/underrated players/ Player types? You could use either player rank or the player score they get(Average player rating). I’d expect it to be OK for offense and miss on the high DRAPM guys.

  6. Andri on February 17, 2013 at 11:37 am

    Not only interesting, but also a necessary analysis.

    It would be great seeing the correlation between RAPM and contracts, as a measure of real under or overrating with consequencea.

Leave a Reply

Your email address will not be published. Required fields are marked *

Current day month ye@r *

DSMok1 on Twitter

To-Do List

  1. Salary and contract value discussions and charts
  2. Multi-year APM/RAPM with aging incorporated
  3. Revise ASPM based on multi-year RAPM with aging
  4. ASPM within-year stability/cross validation
  5. Historical ASPM Tableau visualizations
  6. Create Excel VBA recursive web scraping tutorial
  7. Comparison of residual exponents for rankings
  8. Comparison of various "value metrics" ability to "explain" wins
  9. Publication of spreadsheets used
  10. Work on using Bayesian priors in Adjusted +/-
  11. Work on K-Means clustering for player categorization
  12. Learn ridge regression
  13. Temporally locally-weighted rankings
  14. WOWY as validation of replacement level
  15. Revise ASPM with latest RAPM data
  16. Conversion of ASPM to" wins"
  17. Lineup Bayesian APM
  18. Lineup RAPM
  19. Learn SQL