The cumulative game


There’s something to be said for combining a very low turnover rate with normal (or even below-average) offensive rebounding, a la Villanova. Conversely, other teams may be underperforming by four or even five points per game due to a sheer lack of scoring chances. (Bill Streicher, USA Today Sports)

Over the past couple years, I’ve started wondering whether the manner in which our brains are hard-wired is conspiring with the inherent nature of basketball to keep us from recognizing how important it is to generate a lot of shot attempts.

Consider the Premier League. A proper appreciation of shots that had a chance to go in but didn’t for Arsenal or Chelsea constitutes a rudimentary level of “well, duh” analysis. In that setting, shots on goal are a really big deal. They’re tracked closely and dissected individually by the studio talent after the match.

In basketball, however, attempted shots are the fabric of the game itself. Attempts in this sport are numerous, unremarkable in isolation, and, indeed, most of them (56 percent, give or take) are misses. Shots that don’t go in aren’t exciting visually, they’re never featured in highlights, and each miss represents a failure, of sorts.

All true enough. But where we’re led astray during an actual game, I think, is in our real-time insensitivity to cumulative effects. I plead guilty as charged there, too, but lately I’ve been trying to do better. Specifically, I’ve had to affirmatively alter the manner in which I watch games involving Creighton or Virginia Tech, among others.

I’ll return to the Bluejays and the Hokies in a moment, but first allow a note on the path traversed so far….

You can’t get an offensive board if you’ve already committed a turnover
Two years ago I let my enthusiasm for a new stat get out ahead of my skepticism, and I wrote a rather sweeping headline to the effect that Final Four teams are better at taking shots than they are at making them. Then self-recrimination set in.

Searching for a simple and easy-to-remember item that would at one stroke hack this whole shot volume thing the way Drew Cannon’s “easy bubble solver” beautifully hacked bracketology, I instead ended up with something that went too far toward treating turnover-less possessions and offensive boards as coequal. They’re not. They are instead sequential, the former being a necessary but not sufficient condition for the latter. Pondering that fact, I made a Scarlett O’Hara-style vow that, as Dean Oliver is my witness, I’ll never be hasty again.

In an attempt to do better, I discarded the “easy to remember” part of my old mission statement entirely and adopted a new, rather more unwieldy, but also more accurate shot-volume metric, one that correlates better to points per possession than did the old one. While I was at it, I also added more teams to the data set, and — hey, what do you know?

The enthusiastic headline looks pretty good after all….

Final Four teams are better at taking shots than at making them

Standard deviations above conference mean
Final Four teams, 2011-18
Conference games only
             Accuracy (eFG%)      Shot volume (SVI)
SD's              +0.88                 +1.19

Maybe the Premier League comparison’s not so far off. Just as a first-level approach to summing up a given team’s performance, I now find I’m often looking at: 1) defense; 2) shooting accuracy; and 3) shot volume, where the relative importance of all of the above is ranked accordingly. Defense is half the game, and accuracy’s more predictive of scoring, of course, than is volume. But volume’s more predictive of scoring than are either turnovers or offensive boards alone.

For one thing, such an approach puts any team’s turnover and offensive rebound rates into context. To take one example, Villanova’s offensive rebound percentage in Big East play last season ranked eighth in the conference, which sounds pretty bad. Then again, the Wildcats recorded the best shot volume in the league, so why sweat the number for offensive boards?

Shot volume index (SVI)
Turnover percentage, offensive rebound percentage, and shot volume
Big East games only, 2018

                         TO%     OR%     SVI
1.  Villanova           13.8    25.2     98.2
2.  Seton Hall          16.6    31.4     97.7
3.  Butler              14.6    25.6     97.4
4.  Xavier              17.2    29.4     96.1
5.  Providence          17.0    28.6     96.0
6.  Marquette           17.1    26.7     95.1
7.  St. John's          16.2    19.6     92.8
8.  Creighton           16.0    18.8     92.7
9.  DePaul              21.5    32.8     92.6
10. Georgetown          21.0    29.0     91.5

Nominally poor offensive rebounding wasn’t a problem for Jay Wright. How can we know when a high turnover rate and/or a low offensive rebound percentage is a problem?

Let’s stick with the 2018 Big East. In league play, Creighton had, effectively, the same shot volume as DePaul. For their part, the Blue Demons had the worst turnover rate in all of major-conference basketball in 2018, even worse than what Pitt was recording in the midst of a once-every-few-decades meltdown.

Perhaps that’s interesting analytically.

An extreme deemphasis on offensive boards simulates turnovers you don’t commit
Caution is required, of course, when speaking of a counterfactual like how many points a team “should” have scored.  For starters, we have to reckon with the surprisingly conditional and not necessarily linear relationship between accuracy from the field and scoring. If you’re bad at both, your scoring will tend to be statistically better than your shooting, but, if you’re good at both, your shooting will tend to be statistically better than your scoring.

This dynamic explains why, if we derive a cumulative 13-year translation rate from every major-conference team-season between overall accuracy (49.5 eFG%) and overall scoring (1.04 points per possession), we end up imposing a highly unrealistic evaluative burden on the best shooting teams. In fact, by these lights, even the historically great Villanova offense from last season scored fewer points than it “should” have. To me, that feels like one counterproductive benchmark.

I’ve therefore found it more useful to treat the most accurate teams as their own discrete subgroup, one with its own group-derived accuracy-to-points expectation. Call this our “high-accuracy translation rate.”

Applying our more forgiving high-accuracy translation rate to last season, we find that Creighton’s low shot volume in Big East play would be projected to have cost the Bluejays something on the order of six points per 100 possessions. Even that number, however, falls short of the seven points per 100 trips that Virginia Tech may have left on the table in ACC play in 2017 due to even-better-than-Creighton-in-2018 shooting being muffled by even-worse-than-Creighton-in-2018 shot volume.

This is how the Hokies were able to record the best shooting any team’s posted in that league’s games in at least 15 years while at the same time finishing as merely the No. 7-ranked offense in the 2017 ACC. Creighton’s extremes weren’t quite that extreme last season, but this kind of historically aberrant relationship between accuracy and points is, arguably, the dispositive performance characteristic being exhibited by both offenses at the moment.

Accurate but starving (ABS)
There’s zero “wrong” with a 19 percent offensive rebound rate in league play per se. College teams that choose to heavily deemphasize offensive boards are being penalized by what is, as root, a social dynamic, and that feels less fair and definitely less worthy of corrective X’s and O’s than a true basketball performance imperative.

In effect, such college teams are being hurt by offensive rebound percentages that would look rather more normal in the NBA. The problem is these teams aren’t in the NBA.

Virginia Tech, for example, is playing in the ACC, where opponents are generating far higher shot volumes. Indeed, the opponents enjoying lopsided advantages in scoring opportunities over the Hokies include not just extreme outlier North Carolina, but normal everyday Miami and NC State working stiffs as well.

Yes, floor spacing and the number of threes you attempt play their part here, and the issue is therefore remaining in shot-volume contact with other teams that also space the floor and shoot a ton of threes. See Villanova.

Yes, it’s ritually insisted that total doctrinaire deprivation on the offensive glass must surely pay off in terms of transition defense. Fine. Prove it with a correspondingly amazing composite ranking for “percentage of opponent FGAs in transition” (per slash “shot volume,” one even higher than the composites posted by better overall defenses with, like the vast majority of Division I, far better shot volumes. See Virginia.

If I were an assistant, I’d get a smoke detector from the hardware store, put a label on it saying “shot volume sensor,” and press the “TEST” button any time my highly accurate team dropped below the conference mean by more than half a standard deviation. The alarm would signal that my team had officially become accurate but starving.

Teams that earn the ABS label do so cumulatively and therefore in a way that’s exceedingly difficult for us to apprehend fully during games. Even so, such teams face a measurably more difficult hill to climb on offense, no matter how accurate they are.

Bonus fine print. Now the disclaimer. This business of the occasional highly accurate team being penalized for its thoroughgoing devotion to suppressing offensive rebounds is purely an accident of history and not some iron law of basketball.

Actually, the teams we saw falling short of the shot-volume mark a decade ago were typically doing so due to an excess of turnovers. Historians of the game, I give you the O.J. Mayo USC team of 2008, the blip “down year” UCLA team of 2010, and, the piece de la resistance in the field of highly accurate scoring futility, the Byron slash B.J. Mullens Ohio State team of 2010.

True, Washington State performed a rather moving tribute to the high-accuracy slash high-turnover lifestyle as recently as last season. Nevertheless, the team that’s simultaneously accurate and prone to give the ball away has become exceedingly rare over the past few years as turnover rates have plunged across college basketball.

Now the problem tends to be instead a relative lack of offensive rebounds. The irony there is that shunning second chances to a down-in-the-300s-at-KenPom extreme merely serves to exaggerate the otherwise slight significance of offensive rebounding in terms of performance.