You Are Here > Basketball-Reference.com > BBR Blog > NBA and College Basketball Analysis

SITE NEWS: We are moving all of our site and company news into a single blog for Sports-Reference.com. We'll tag all Basketball-Reference content, so you can quickly and easily find the content you want.

Also, our existing Basketball-Reference blog rss feed will be redirected to the new site's feed.

Basketball-Reference.com // Sports Reference

For more from Neil, check out his new work at BasketballProspectus.com.

Is Floor% a Better Predictor of Future Efficiency Than Efficiency Itself?

Posted by Neil Paine on January 6, 2011

I was reading Brian Burke's excellent Advanced NFL Stats site when I came across this post about predicting future team rushing efficiency (expected points per rushing play). Because a handful of big, somewhat unpredictable rushing plays can have such an outsized impact on overall efficiency, Burke found that past success rate -- simply the percent of plays that had positive expected point values, regardless of their magnitude -- was actually a better predictor of out-of-sample rushing efficiency than past efficiency was.

In basketball, we have two similar (though not totally analogous) metrics: Offensive Rating (average points scored per possession) and Floor% (the probability of scoring at least one point on a given possession). Offensive Rating gets all the publicity, and as well it should -- the entire goal of an offense is to maximize points per possession. However, ORtg can also be heavily impacted by 3-point shooting, so boom-and-bust offenses that over-rely on threes might be like those teams whose running backs bust off a handful of long runs but otherwise get stuffed at the line too often. Their overall efficiency might be good, but their success rate isn't, and in the end success rate is what you can count on going forward.

With that thought in mind, I'm going to replicate Burke's study, hoops-style. The NBA's rapidly-increasing obsession with 3-point shooting finally leveled off from 2008-10, so my sample will include every game from those seasons. For those games, I calculated each team's offensive/defensive rating and floor%; I then broke their seasons up into even- and odd-numbered halves based on the order of games in the year, as well as 1st & 2nd halves of the schedule. Finally, I ran the correlation between ORtg/DRtg or offensive/defensive Floor% in a given half and ORtg/DRtg in the opposite half. Here were the results:

Grouped by Half of Season

Correlation between... R
ORtg & opposite-half ORtg 0.569
Floor% & opposite-half ORtg 0.490
DRtg & opposite-half DRtg 0.576
Def. Floor% & opposite-half DRtg 0.573

Grouped by Even/Odd game #

Correlation between... R
ORtg & opposite-half ORtg 0.778
Floor% & opposite-half ORtg 0.674
DRtg & opposite-half DRtg 0.762
Def. Floor% & opposite-half DRtg 0.755

As you can see, efficiency is still the best predictor of out-of-sample efficiency, and that's true on both sides of the ball. What's interesting, though, is that defensive Floor% is very nearly as effective a predictor of DRtg as DRtg itself, while offensive Floor% is nowhere near as effective as ORtg in predicting out-of-sample ORtg. This is because a team's proportion of offensive 3PA/FGA is largely by choice and therefore stays consistent over time. On defense, though, a team faces a wide variety of opponents, each with different ratios of 3PA/FGA. Over the course of a season, the distribution of opponent 3PA/FGA rates is much narrower than the distribution of offensive 3PA/FGA rates.

ShareThis

5 Responses to “Is Floor% a Better Predictor of Future Efficiency Than Efficiency Itself?”

  1. Jason J Says:

    Neil - How is floor% calculated?

  2. DSMok1 Says:

    Do you think you could do a logistic regression model incorporating both inputs in some manner? That way both efficiency and consistency could be incoporated.

  3. Neil Paine Says:

    Re: #1 - Floor % = Scoring Possessions / Total Possessions

    where

    Scoring Poss = FG + (1 - (1 - FT%) ^ 2) * FTA * 0.4

    Re: #2 - Just to clarify, you want to use a logit model to predict out-of-sample floor% with Floor% and ORtg as the inputs?

  4. DSMok1 Says:

    No, I would attempt to predict out of sample winning percentage. Seriously.

  5. BILL WHITE Says:

    WAY TO GO PAL!?????????????????>l<?