Sunday, June 24, 2012

How Does Tempo Impact Efficiency?

It's the offseason, and it's been a while since I've posted. Since there are no games to talk about, I thought I'd address an "advanced stats" issue that often comes up for debate - tempo free stats. Certainly my regular readers are more supportive of tempo free stats than most of the dinosaurs that call games on television or talk about them for major newspapers/magazines, but there's a myth that I wanted to explore about the relationship between efficiency and tempo.

Let's start with a couple of graphs that should make sense - offensive and defensive PPG vs raw tempo (possessions per game). The data below includes all 345 Division I teams from the 2011-12 season. I have thrown a best fit line on the data, with the equation in the bottom right hand corner. Don't worry about the y-intercept - what matters is the slope and the R-squared. If you forgot your high school statistics class, R-squared tells you how well the data fits the "best fit" line. So in this case, an R-squared of 1 would mean that all of the data fit perfectly on that line, and an R-squared of 0 would mean zero correlation (though for those of you that took some college stats, you'll know that for a finite-sized sample, even with no correlation you'd expect a tiny positive R-squared just from pure randomness). Anyway, check out the data below:

The first thing you see on those two plots is what you expect - the more possessions per game the more points per game a team scores and allows. In fact, the data says that for every extra possession per game, teams score 1.15 more points and allow 1.22 more points. And you can see that the fit, while obviously not perfect, is pretty good. Clearly, the faster you play the more points you score and the more you allow - I don't think this is disputed by anybody.

But what is disputed is whether tempo can affect efficiency. It's taken as objective truth by most basketball announcers and mainstream media analysts that it does. Teams that play slow can "grind" a game and "frustrate" an opposing offense, while teams that run can get more fast breaks and wear out opposing defenses. Well, is this true? Let's check out some data below:

What you see plotted above are the Pomeroy "adjusted" offensive and defensive efficiencies vs Pomeroy "adjusted" tempo from the 2011-12 season. For those unfamiliar with efficiency, the numbers on the y-axis are points scored/allowed per 100 possessions. The best-fit lines for this data suggest that there might be a tiny correlation between higher tempo and more scoring, but probably not - the R-squared values are basically zero. And some further data backs that up:

The first two plots above are the raw offensive and defensive efficiencies from 2011-12 plotted vs the raw tempos, in case you don't trust the Pomeroy stats. Though it turns out that this really didn't change anything. And below that I've plotted the adjusted offensive and defensive efficiencies plotted vs adjusted tempo from the previous season (2010-11). In that case, there's actually a slight negative correlation between tempo and offensive efficiency.

In other words, the myth is wrong. Faster tempos do not make it easier to score, and slower tempos do not make it easier to defend. There are some very slow elite defenses (Wisconsin and Virginia being the most obvious examples from this past season) but there are also very quick elite defenses (North Carolina and Marquette come to mind from this past season). At the same time there were very quick elite offenses (Iona) and very slow elite offenses (Florida).

What's really fascinating to me is what I saw when I plotted Pomeroy efficiences vs point per game stats:

Naturally teams that are more efficient offensively score more points and teams that are more efficient defensive allow fewer points - that's pretty obvious. But compare the best fit lines to the two plots at the top of this post, which are PPG vs tempo. Scoring efficiency correlates with PPG a little bit better than tempo... but only a little bit. So if a team scores a ton of points per game, it tells you something about how good they are at scoring, but almost as much about how fast they play. Announcers during games routinely tell us that one of the teams we're watching is really good at scoring because they score a lot of points per game, or are really good at defense because they allow very few points per game, but it turns out this is only half the story. PPG tells us almost as much about tempo as it does about how good a team is on offense or defense.It's possible for an elite offense to score 65 ppg and for a poor offense to score 70 ppg - it happened this past season!

As an aside, staring at this data got me thinking about tempo vs team quality - do the best teams tend to run or tend to slow things down? Let's look at the data first:

What the data says is that there is simply no correlation between tempo and Pomeroy rating. Elite teams are just as likely to play really quickly as they are to play really slowly. In a vacuum, this seems like a market inefficiency. Bad teams should prefer slower games and elite teams should prefer faster games. The greater the number of possessions in a game, the less randomness can play a role. A bad team has a much better chance of winning a 50 possession game than a 90 possession game against a superior opponent.

Obviously some coaches have certain systems that they've honed over years, and these systems require fast or slow play. Force a Mike Anderson team to grind the ball and they won't play as well - same goes for asking a Mike Brey team to run. But for coaches that are more flexible in their system, or that are starting fresh, it might be worth considering a quicker or slower tempo depending on your team quality.

But if there's one takeaway from this post, hopefully it's that you get as irritated as I do when announcers tell us that PPG tell us how well a team plays offense or defense. If you're an analyst: ignore tempo and tempo-free stats at your own peril.


Adam said...

Good post! I'm not sure about the market inefficiency point though. The key words you use are "in a vacuum," because I don't think it's true that all efficiencies will remain equivalent if you adjust the tempo; in fact I'm pretty sure they would not.

There are reasons for good teams to slow it down. One I can think of off hand is to best utilize interior height and offensive rebounding ability. And the converse is a reason that bad teams may push the tempo when possible. Sometimes a transition 3 is a much better offense than getting an suffocated for the full shot clock.

Lots of tradeoffs in play.

Jeff said...

I agree. Certainly if your offense revolves around an unstoppable big guy in the post then you don't want to fast break all the time.

I'm just arguing that on the margins, you'd imagine that programs that know they can't bring in great talent would try to slow things down.

In fact, I've seen that in person. I've been at November/December "buy" games where some MEAC or SWAC team would dribble the ball up the court and stand in place for 20 seconds before running the offense just to try to shorten the game.

But like you said, it's about style. If you are coaching at a SWAC school and your whole life you've taught full court press... well, I guess you have to stick with what you're good at.