Basketball Prospectus: Unfiltered Everything Else is Fluff.

February 15, 2012

Using B-R.com’s New Feature to Study Late-Game Situations

Filed under: Uncategorized — Kevin Pelton @ 4:27 pm

It’s the day after Valentine’s Day, but Basketball-Reference.com has provided a terrific gift to those of us who love the site. Today, Justin Kubatko unveiled Play Index+, which makes play-by-play data searchable and offers shot charts and location data. The list of studies that can be performed with this tool is virtually inexhaustible, and I look forward to using it repeatedly in the months and years to come.

For now, though, one particular thought came to mind. Play Index+ offers us the ability to study clutch shots in a much more robust fashion than we have in the past. Specifically, I went back to a 2009 Unfiltered post on shots to tie or take the lead at the end of the game, in which I tried to answer the question of whether role players fare better than stars in these situations but was limited somewhat by the available data. Not so anymore.

I isolated shots to tie or take the lead in the final 30 seconds of the fourth quarter and overtime from last season. There were 610 such shots, on which the NBA as a whole shot 29.0 percent. That includes desperation heaves, but the number is still stunningly low. Even on two-pointers, the league shoots just 36.6 percent in clutch situations. Why that number is so low is part of the question.

One theory, espoused frequently by TrueHoop’s Henry Abbott, is that offenses suffer from being too predictable and relying on isolations for their star players rather than balanced offenses. We can’t entirely test that theory, since we don’t know what led to each shot, but we can compare how players with many attempts to tie or take the lead performed as compared to those with few attempts, presumably role players getting a feed from a star to take an open shot.

Here’s how the statistics break down by the number of attempts the player had:

Att  FGA    FG%    2P%    3P%   eFG%  
10+  36   .222   .222   .222   .250
9    45   .378   .462   .263   .433
8    56   .321   .425   .063   .330
7    63   .333   .423   .270   .413
6    48   .250   .259   .238   .302
5    75   .280   .317   .235   .333
4    56   .250   .333   .154   .286
3    81   .309   .447   .186   .358
2    86   .267   .400   .122   .297
1    64   .281   .333   .226   .336  
Tot 610   .290   .366   .199   .335  
5+  323   .300   .353   .228   .348
4-  287   .279   .384   .170   .321

Even when we aggregate by number of shot attempts, the sample sizes are still tiny and the data noisy. Still, we distinctly do not see the pattern we’d expect if teams go to their stars too frequently. In fact, because of their superior three-point shooting, players with many attempts to tie or take the lead tend to be slightly more efficient than players with few of them. Mostly, the numbers tend to show that the attempts are more or less being distributed as they should. In a world with perfect game theory, the percentages for all levels of attempts would end up the same. We’re apparently not as far from that as you might think.

No Comments

No comments yet.

RSS feed for comments on this post.

Sorry, the comment form is closed at this time.

Powered by WordPress