The Playoffs in a Make or Miss League – What IS A “Good” Shot?

May 5, 2015; Oakland, CA, USA; Golden State Warriors guard Klay Thompson (11) shoots the basketball against Memphis Grizzlies forward Zach Randolph (50) during the second quarter in game two of the second round of the NBA Playoffs at Oracle Arena. Mandatory Credit: Kyle Terada-USA TODAY Sports

A major part of the appeal of the NBA Playoffs is the way game-to-game adjustments allow for each series to take on a real narrative arc. The effects of each move and countermove can readily be judged, and teams which make the proper adjustments move on, while those which fail to do so go fishing.

Except it’s not that simple. In any one game sample, a team might miss even the easiest of shots at astonishing rates, while the other team might have a player scoring time after time despite defenders draped all over him. There’s a tendency to see the results and say “that guy killed us, let’s change up our coverage” when all that was really needed is for another game to start and the laws of probability to punch those contested pull-up jumpers in the face. On the flipside, a team getting the looks they want and just missing might not need to change much at all – as the adage goes, “it’s a Make or Miss League.” Meaning some nights the ball doesn’t go in and you lose.

Discerning the real problems from short-term variance[1. The long run can take a long time to arrive. On 25 occasions this regular season, a team had a seven game stretch where they shot more than 10% worse on aggregate than their season average on open threes. These 25 instances occurred over 11 separate stretches, the longest of which was Indiana’s horrific run over most of January. In other words, there is about a %1 chance a team won’t be able to throw the ball in the ocean over any current seven game stretch.] is always an issue, and human nature is to react more to results than process. A non-shooter hits a long jumper or two and suddenly defenders are flying by him on pump fakes when “let him keep shooting” remains the best plan.

Similarly, a player who has been missing is expected to keep missing; there was a moment late in Game 2 of the Hawks-Wizards series when Marcin Gortat almost disdainfully backed off Al Horford on the baseline. Horford has been cold from midrange, but is a highly proficient from those areas. Horford calmly drilled the 18-footer, and before running back on defense appeared to turn and say something to the Wizards bench, presumably something along the lines of “come on now, you know I’m not going to keep missing those.”

At the same time “we’re just missing shots” can become an excuse and prevent a team from casting a critical eye to things which might in fact need to be adjusted. Critical in making that distinction is an honest reckoning of shot quality. Thankfully, we’ve come a long way from pure eye-test measurements. For one, we’re talking aggregates, and it’s almost impossible to fully contextualize and evaluate the 150 or more shots without becoming hopelessly biased by the end results. In real time, the vast majority of shots look like “better” shots when they go in. Another difficulty is the viewer’s perspective, whether on television or in the arena, often makes it challenging to distinguish a shot which is actually and effectively contested versus a hopeless leap from 10 feet away or a “fake hustle” late closeout.

To some degree, SportVU data helps with the latter problem – we know now exactly how far shooter and closest defender were[1. Center-mass to center-mass. Objects in SportVU are usually about three feet, the length of an arm, closer than they appear.] at the time of release. While this is still imperfect – whether a hand is up or down isn’t tracked by the cameras, it’s unknown if the defender is in the shooter’s eyeline or not, and the speed or lack thereof with which the defender is running out isn’t captured.[2. At least in the public data, it’s surely derivable in the full data set.] Imperfect is far superior to nothing in this case, however. It’s not difficult to get a rough understanding of what proportion of a teams’ shot attempts were, broadly speaking, contested.

In fact, NBA.com tracks this on a game-by-game basis:

The above is from the Grizzlies stirring Game Two victory over Golden State at Oracle. While Memphis undoubtedly played well defensively, there are some questions as to how sustainable the effects will be, as Golden State appeared to miss a lot of open shots. The 7-24 “uncontested” shooting appears to bear that out. The interpretation isn’t quite so straightforward, however. In these box scores, a contested shot is defined as one taken with the defender within 3.5 feet[5. Closer examination of the underlying data suggests it’s actually measuring if the defender is within 4 feet rather than 3.5, but that only changes the analysis at the margins.], or just over arm’s length.

It’s a decent first-pass indicator. There is much more to look at, though. The most important additional factor is that the level of defensive pressure is heavily linked with the distance of shot. This makes intuitive sense, as Day One defensive principle number one is protect the basket first. Still, it’s useful to have the degree of this correlation illustrated:

The shots most likely to be contested are right at the basket. But that’s ok, because even a contested shot near the basket is, generally speaking, a great shot! A team taking a large number of contested shots might have been exceptionally well-guarded. Or they may have gotten to the rim at will. On the other side, not all uncontested shots are created equal either. A sizable majority of jump shots, whether midrange or from outside the arc aren’t “contested” by this definition, but the value of these shots differs dramatically:

The degree to which three pointers are more valuable than midrange shots is massive, and the difference in overall shot quality between the uncontested shots being threes or not is similarly huge.

Further, there are open shots and the open shots. Looking at just three pointers[3. Note that only around 2.5% of threes are attempted a defender 2 feet or closer, so the far left of the chart is dealing with some tiny sample sizes.]:

Shots with the defender 4 feet away are “uncontested” just as if the defender were 10+ feet away. But the difference in average accuracy is the same as the gap between C.J. Miles and Danny Green as shooters. A difference in 24 points per 100 shots is kind of a big deal. Especially in a close playoff series when those theoretical quarter points add up quickly.

As should be readily apparent, one can go as deep down the rabbit-hole as one wishes in terms of breaking down shot quality. Other factors such as the identity of the shooter, game & shot clock time, current margin, fatigue, number of dribbles before the shot, precise floor locations and angles and so on can all be taken into account. But with sample sizes already dangerously small, increased detail can actually exacerbate the uncertainty of any estimate of shot quality. While practitioners in the league can figure the exact best point for this analysis for purposes of high leverage decision-making, for our purposes as fans something simpler will do. Horford continuing to miss wide open midrange shots would be a surprise as he shot 52.7% on those shots during the season. Unless of course his dislocated finger is still bothering him. But that’s another question rather than answering whether the looks are good or its time to make changes.

To circle back around to the Warriors, Klay Thompson and Steph Curry combined for 4-13 on uncontested shots. While Curry missed one “alone in the gym” three[5. At a fairly critical moment, momentum wise it must be added.], Thompson had defenders, most notably Tony Allen, in reasonably close proximity on all his shots. While Klay might expect to make one or two more, a more encouraging stat for supporters of the Grizzlies is the fact that Golden State was limited to 24 uncontested shots after taking 39 in Game One which was right around their season average.

So upon reflection, the Dubs did miss some shots, but credit is probably due to the gritty defense of Memphis more so than bad shot luck. It might be tempting for Steve Kerr to figure the Warriors will shoot better next game, some adjustments may in fact be needed to get some better and more open looks for his team.