Series Articles:

Introduction
Part 1/4: Rodman was a better rebounder than you think
Part 2/4: Rodman’s rebounding was more valuable than you think
Part 3/4: Rodman was a bigger winner than you think
Part 4/4: Rodman belongs in the Hall of Fame [or not]
Miscellaneous

 

Summary of Main Points:

Post

Rodman-Specific

Broader Analytical

Part 1(a): Rodman vs. Jordan

  • Dennis Rodman has dominated Rebounding Percentage more than anyone has dominated any major stat.
  • Before Rodman, we should have expected a rebounder of that quality to appear about once every 400 years.
  • Use standard deviations to measure relative greatness
  • Outliers can skew their own data against them

Part 1(b): Defying the Laws of Nature

  • Rodman had an almost unnatural ability to dominate rebounding on both ends of the court simultaneously
  • Rodman showed no trade-off between offensive and defensive rebounding rates
  • The trade-off between offensive and defensive rebounding exists, and is generally a completely separate phenomenon from rebounding ability.

Part 1(c): Rodman vs. Ancient History

  • Contrary to popular opinion, Rodman was a much better rebounder than Wilt Chamberlain or Bill Russell, and it’s not close
  • Total Rebound Percentages for players prior to 1970 can be estimated with extreme accuracy

Part 2(a)(i): Player Valuation and Conventional Wisdom

  • On induction, Rodman will be the worst scorer and the best rebounder in the Hall of Fame
  • Rodman scored even less than we would expect the lowest scoring Hall of Famer to have
  • Rodman was an even better rebounder than we would expect the best rebounder in the Hall of Fame to be
  • The Hall of Fame likes point-scorers better
  • Everyone uses statistics, yet no one listens to statisticians—in part because statisticians build overreaching models, then believe and defend them
  • Rebounding percentage correlates more strongly with winning than points per game.
  • Added: Individual rebounding percentage has a more causative effect on team rebounding percentage than individual PPG does on team PPG.

Part 2(a)(ii): Player Valuation and Unconventional Wisdom

  • Player Efficiency Rating ranks Dennis Rodman as the 7th best player on the 1995-96 Bulls championship team
  • Player Efficiency Rating is terrible
  • Extreme rebounding ability like Rodman’s may have exponential value
  • Player Efficiency Rating fails completely as a predictor of true player value
  • PER rewards Usage rate (shooting), despite no correlation between Usage and shot efficiency
  • PER’s many layers of complications and adjustments are demonstrably counterproductive

Part 2(b): With or Without Worm

  • Rodman has the highest Margin of Victory differential of any player since 1986 with a remotely similar sample size
  • Rodman’s value comes mostly from extra possessions from extra rebounds
  • Despite claims that he was exclusively a defensive player, Rodman’s teams played significantly better on offense with him in the lineup, even after accounting for his offense rebounding
  • Introduce my brand of game-by-game “With or Without You” stats
  • Two main areas of player impact: Reciprocal Opportunities, and Reciprocal Efficiency
  • Impact on one aspect of a game can easily be reflected in other areas statistically
  • Completely independent confirmation of the results of previous analysis is powerful evidence

Part 3(a): Just Win Baby (in Histograms)

  • Rodman’s Win Percentage differential is even better than his Margin of Victory differential
  • Specifically, his Win% differential is #1 of the 470 players who qualified for the study—by a wide margin
  • Adjusting for the quality of teams Rodman played for makes his differential even better
  • Introduce win percentage differential, which is incredibly useful for research and hypothesis-testing in many contexts
  • It is harder to have a big impact on better teams
  • In a game of small margins, exceptional performance in limited areas can be more valuable

Part 3(b): Rodman’s X-Factor

  • Not only is Rodman’s Win% differential greater than his (already great) MOV differential, it is greater by one of the largest margins of any player
  • After adjusting for sample size, Rodman’s X-Factor is by far the largest of any qualifying player
  • The most plausible explanations for this disparity suggest that, in Rodman’s case, his Win% differential may be the more trustworthy metric
  • “X-Factor” is the difference between MOV-predicted and actual Win% differentials
  • Higher MOV’s and larger sample sizes should correspond to smaller X-Factors
  • 3-D plots are a visually appealing way of identifying less-obvious outliers
  • There are several plausible causes for real team and/or player X-factors

Part 3(c): Beyond Margin of Victory

  • Using a standardized model for combining MOV and Win% differentials (which weights MOV more heavily), Rodman still places #1 in the set of 462 qualifying players (many of whom have MUCH smaller samples)
  • Depending on which metric you favor, Rodman’s differentials place between the 98th percentile and the 99.98th percentile among full-time players (approximately 5% make the Hall of Fame)
  • The statistical community over-values Margin of Victory and under-values raw winning percentages
  • Winning is a provably existent skill, separate from scoring and allowing points.
  • Predicting regular season win expectations is best done with a combination of Margin of Victory AND Win percentage
  • The larger the sample size, the more heavily Win % should be weighted

Part 3(d): Endgame: Statistical Significance

  • Rodman’s win differentials alone are statistically significant well beyond the 99% level of confidence.
  • Looking at the overall statistical significance for player win differentials over the broadest possible pool of 1539 players, Rodman ranks between 2st and 8th, depending on your preferred metric.
  • Rodman’s average ranking across metrics is second only to Shaquille O’Neal (who’s sample includes over twice as many qualifying games).
  • Introduce “Black Box” as a term for when variance gets eaten up by events that have binary outcomes.
  • Standard deviations for win differentials can be found by sampling
  • These can be adjusted to different sample sizes mathematically
  • Using this process, we can measure the statistical significance of individual win differentials for many players who didn’t have sufficient samples to qualify for earlier comparisons

 

 

Part 4(a): All-Hall?

  • There are indirect reasons to believe Rodman’s Win Differential is more reliable than his Margin of Victory Differential
  • Occam’s Razor and Bayes Theorem reasoning can make unlikely or only mildly supported independent hypotheses much more likely

Part 4(b): Rodman v. Jordan 2

  • Rodman’s unusual consistency and otherworldly impact per possession used suggest that he may have had even more value than his Win Differentials indicate
  • Introduce “Invisible Value” and “I-Factor”

 

Selected Visuals:

1/4(a) Rodman v. Jordan

1/4(b) Defying the Laws of Nature

1/4(c) Rodman v. Ancient History

Wilt and Bill_28675_image009

2/4(a)(i) Player Valuation and Conventional Wisdomimageimage
2/4(a)(ii) Player Valuation and Unconventional Wisdom

image_thumb13_thumb

image4_thumb3

2/4(b) With or Without Worm

image

image
3/4(a) Just Win, Baby (in Histograms)
image3/4(b) Rodman’s X-Factorclip_image005clip_image0063/4(c) Beyond Margin of Victoryimage12_thumb1image_thumb163/4(d) Endgame: Statistical Significanceimage5_thumb4/4(a): All-Hall?

4/4(b): The Finale (Or, “Rodman v. Jordan 2″)

image_thumb21

Rodman Teams

image_thumb19

Graph of the Day: Rodman, Visualized—An Outlier in Motion

20 Responses to “The Case for Dennis Rodman: Guide”

  1. Matt says:

    Awesome, I had forgotten how much information was in the earlier posts, great series all around. What can we be expecting next?

  2. Eric says:

    I’m glad to see that our conversation from the firm holiday party made it into not one, but several blog posts. I wonder how Kevin Love’s stats this season — closing in on the consecutive double-double mark, leading the league in rebounding (per game and per 40) — compare to those of the Worm. Though Love’s accomplishments are pretty impressive, maybe he is just stealing rebounds from his teammates. And he sure can’t be contributing much in the win column.

    • benjaminmorris says:

      Hi Eric,

      I actually had a long conversation about Love at the SSAC yesterday. He definitely seems to be having a statistically rare season. The biggest problem with him is that, in the NBA, great players pretty much don’t lose ever. He’s going on his third straight season with a winrate of <30%. The only marquee players to do this in any seasons over the past 30 years are Dwyane Wade (who did it once) and Pau Gasol (twice).

  3. [...] Rodman owns by a large margin the highest career rebounding rate, a metric he dominated more thoroughly than Jordan did scoring. Though I’m admittedly out of my element here. Smarter people than me or Pesca have hashed out the Rodman debate, and if you fashion yourself one of these, I’d direct you to this labyrythian minefield of numbers. [...]

  4. Benjamin says:

    Hi guys, new to the site. How would you respond to the assertation that RB% does not work for “role players” (can be defined as players who don’t play the most on their team) since they can expend more energy on rebounds/get more rest, etc. Rodman didn’t necessarily play the number of minutes you’d expect from an elite player; that said, he still compiled otherwordly rate-stats (RPG, RP40, etc). Do you feel RB% has this limitation?

  5. [...] truly wonderful statistical analysis making a case for Dennis Rodman. One of my favorite lines: Before Rodman, we should have expected a rebounder of that quality to [...]

  6. [...] is a case for Rodman as the greatest player in NBA history.  It’s worth reading it all. It has a dozen different charts, measuring different aspects of [...]

  7. [...] is being misrepresented (and have a few days to spend being entertained) go read Ben Morris’ Rodman Opus and don’t miss the bit about rebounding being [...]

  8. [...] related. His value as a baller is a great topic, and to learn more about that I direct you to a series of articles far better than anything I have ever written. This article, though, will mainly focus on [...]

  9. dq says:

    I’ve looked at this (but not read every single word) but how do you go from his point margin difference with and without, which is barely ahead of others, and come back with the incredible high winning %?
    Basically I agree through 3A in the chart above, but then I come to a leap that I don’t get (yet)

    thanks

  10. dq says:

    I’m with you through 3A. How do you go from a point margin barely better than number 2 to a winning % far better than anyone else?
    That’s my problem with this so far, and the leap I want to understand

    thanks

    • The Win% differential is calculated separately using the method described. It isn’t projected based on points, it’s the actual difference in his teams’ winning percentages with and without him.

  11. dquinn says:

    Rodman’ benefits greatly from a .381 differential from the 95 pistons, which had the 20 game minimum you used.

    Had you used 21 games for your cut-off (1/4 of the season, rounded up) you would have different results.

    • Fair enough, but I don’t think the point is particularly probative: His differential was high with every team he played for, and can drop a data point or two and still be one of or the best ever.

      I don’t have the data in front of me, but to dampen outlier seasons for whatever reason (such as age, or, incidentally, a better example is Ron Artest, whose suspension after the “Malice in the Palace” coincided with many team-mates and inflates his career differential substantially), I tried dropping each player’s “best” and “worst” impact years, and it had little effect that would be relevant to this analysis (aside from dramatically shrinking the pool of qualifying players).

  12. dquinn says:

    You also get MJ who has only one qualifying season – 2002 Washington after 2 retirements where is differential is still .182.
    His return season 1995 had 17 games and he had a differential of .242 – if I weight those 2 you get .208 – and neither season was his peak.

    This is such a cherry picking exercise that it doesn’t really mean a lot.

    • Meh, you have to draw a line somewhere. I was acutely aware of the fact that adjusting the filters a small amount one way or another would include or exclude certain players and/or be more or less favorable to them. So I tried to set lines that were general, rationalizable on their own terms, and as effective and probative as possible with my particular goal in mind: evaluating Dennis Rodman. There’s a constant trade-off between breadth and accuracy, and I went with a balance that I thought best for testing my hypothesis about Rodman having HOF-level value.

      If you read the last post in this series, you know that I agree MJ’s Win % diff is probably higher than Rodman’s (though, given sample sizes, it is not more statistically significant). I have also noted in several different places that I think Rodman has certain advantages in these comparisons: For example, he missed a relatively high percentage of games due to suspension rather than injury, and he didn’t stay around in the league several years past his prime as many other great players did. But the fact that there are large margins of error is offset by Rodman’s statistically extreme performance. The whole point of showing that he comes out on top in this reasonably broad pool of players isn’t to show that he is actually the best, but to show how unlikely it is that he wasn’t very very good.

  13. dquinn says:

    well put statement. Good job

  14. Michael Jordan says:

    I’m not impressed.

  15. Mike says:

    In the “Outlier in motion” graph, Rodman seems to make a huge statistical leap in rebounding prowess between the 1990 and 1991 season. What’s happening there? Did his MOV or win% diff also change noticeably (assuming we have a large enough sample size)?

    If we’re looking for the bizarre loophole, maybe looking at other examples of large statistical leaps from year to year+1 would yield worthwhile insights.

    I also wonder if Rodman can recall any conscious change in his approach or strategy for games during those years. Worth asking him before he gets locked away in North Korea? :)

Leave a Reply

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

*

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>