At some point in the distant past, time of possession became a popular stat in football. Control the ball and you win the game. Eventually, analytically-minded people realized that wasn’t true at all. You controlled the ball because you won the game. A lead added incentive to run the ball more often which drained the clock. But time of possession still persists as a staple of the football box score.

There hasn’t been much work in this area on the basketball side. The only thing I’ve seen related to time of possession in the college game was this piece from David Hess from two seasons ago. Yet time of possession can give hoops fans useful information about a team’s style.

One of the most simple and useful things to come out of advanced hoops stats is the calculation of a team’s pace. It’s best use is to give us some description of a team’s style and it also helps us understand how counting stats can be skewed by the opportunities a team has on offense and defense.

It’s also helpful in coaching-change season, when we have a reference for every newly-hired coach that wins the press conference by promising an uptempo style. Even during the season, as we’re assured by someone covering the game that coach so-and-so likes to run when possible, we can look at a team’s average pace and get an idea of whether that’s true.

However, even in cases where a team owns a slow tempo, there’s always been a possible excuse. Maybe the team really does like to run, but their defensive style causes their overall tempo to be slow. One way to test this is to split a team’s pace into offensive and defensive components.

The measurement for component pace is about as simple as it gets: each team’s average possession length (APL, measured in seconds) on offense and defense. The number is not adjusted for competition and it measures the length of an entire possession, so offensive rebounds count. In the future this will probably change, but the raw data is insightful on its own. 

In working on this, I was concerned whether the underlying play-by-play data would be good enough to support such calculations. Anyone that has done work with play-by-plays understands the occasional inaccuracies that can occur. However, I was reassured when checking if there was continuity in a team’s computed APL between seasons.

Here’s a comparison of each D-I team’s offensive APL from 2012 and 2013, and likewise the relationship between defensive APL over the last two seasons.

Not surprisingly, there’s more continuity in a team’s offensive APL than its defensive APL. And as you also would expect, there’s more variation among teams on the offensive side, because the offense possesses the ball and thus is in more control of possession length. But back to that in a moment. It looks like there’s season-to-season continuity in the statistic, but a confirming reference point is season-to-season continuity in a team’s overall pace.

Offensive APL has slightly a slightly better correlation from year-to-year than raw tempo, which is generally accepted as a fairly reliable measure of a team’s pace. In reality, tempo doesn’t exactly capture what a team is trying do offensively. It’s influenced by what’s happening on defense, which by the way, has no relationship to a team’s offensive style.

However, because there’s more variation in offensive APL, tempo is more influenced by that side of the ball, which is demonstrated in the following plots. If all you have is a team’s tempo (which is all we have prior to the 2010 season), you are getting more contribution from a team’s offensive style than its defensive style in determining that figure.

I think what we are seeing here should be expected if the APL data is reasonably accurate. Defensive APL is going to be more noisy because the offense has more control over possession length. If Bo Ryan wants to stall the hell out of the ball, as he did in 2012, the defense is largely at his mercy. Still, the defensive APL is not completely random. It was nice to see Eastern Michigan and Syracuse show the same tendency towards forcing longer possessions since they play an identical and distinct defense. And across Division I, zone and pack-line teams tend to force longer possessions.

Also comforting for analytical purposes is that two of the top three movers in defensive APL last season, Morehead State and South Carolina, had new coaches. Defensive APL will be more jumpy from season-to-season even when there isn’t a coaching change, but it’s good to see that there is some signal there.

There are infinity interesting things to find in the data on a team level, but here a few I noticed…

Of all the coaches in D-I, Jim Boeheim has the biggest gripe about possessions-per-game being an inaccurate measurement of his offense’s style. The last four seasons, the Orange have had an offensive APL rank of 41, 12, 133, and 4, while their adjusted tempo has ranked 252, 209, 177, and 41.

BYU led the country in shortest offensive possessions the past two seasons, but also had longer than average defensive possessions. That difference was more exaggerated last season as their use of the zone increased. They were first in offensive tempo and 325th in defensive tempo. As shown above, there’s more spread in offensive APL and so BYU still ended up with a fast overall tempo anyway.

The 2010 season was interesting – VMI and Seattle ranked first and second, respectively, in both offensive and defensive APL. I wish the old Duggar Baucom would come back.

I can sense you’re suffering from scatterplot fatigue, but here are two more, relating defensive APL and defensive efficiency.

In general, longer defensive possessions make for a better defense. No doubt this relationship can be explained by the tendency to prevent easy buckets in transition and more bad shots late in the shot clock. Stephen F. Austin was the best counterexample last season, regularly forcing short and empty possessions. But they also demonstrate the amount of noise in this measure. Previous seasons reveal a defensive APL around average.

A total of eight teams have ranked in the top 50 in adjusted defensive efficiency and defensive APL over the past four seasons, while 62 times a top 50 defense has ranked in the bottom 50 of defense APL. Kudos to Danny Kaspar and Dana Altman last season, but faster possessions and a great defense do not mix. Here’s guessing that Texas State and Oregon experience defensive possession length that is close to, or even above, the national average next season.

You might see why there’s some incentive to play zone or pack-line defense. I’ll put something together on the continuing decline in scoring when people start coming back to the site, but I’m skeptical that it’s because coaches are control freaks who insist on calling set plays instead of letting their teams get easy buckets in transition. Many coaches are doing more things to prevent opponents from shooting quickly.

In the meantime, APL data is now available on each team’s scouting report going back to 2010. It should add another layer to our analysis of a team’s style. Possessions per game is still the way to determine a team’s efficiency, but if a coach insists he likes to run, his offensive APL is your personal lie-detector test.