Shot Quality, Explained: Separating Shot Selection From Shot-Making

When a player shoots 45% from the field, we know what happened — he made 45% of his shots. What we don't know is whether that was a good night or a bad one. A player who took nothing but wide-open layups and made 45% shot terribly; a player who made 45% on a diet of tightly contested pull-up threes shot brilliantly. Field-goal percentage tells you the result but says nothing about the difficulty. Shot quality is the tracking-era attempt to fix that — to estimate how hard each shot was before it went up, so we can finally separate the shots a player chooses from how well he actually shoots them.

The problem shot quality solves

Every shooting stat we've had until recently — field-goal percentage, effective field goal percentage, True Shooting — is a measure of outcomes. They count makes and misses and weight them by point value. None of them know anything about the circumstances of the shot. A wide-open corner three and a heavily contested heave at the buzzer are both, to these stats, simply a three-point attempt that either went in or didn't.

That's a real blind spot, because two players can post the same shooting numbers while doing completely different things. One generates easy looks and converts them at an expected rate; the other manufactures difficult shots and makes them anyway. Conventional stats can't tell these apart, and so they can't answer two questions that matter enormously: Is this player getting good shots? and Is this player a good shooter, independent of the shots he gets? Shot quality exists to pull those two threads apart.

What tracking-based expected field-goal percentage measures

The core idea is to build a model of expected field-goal percentage — the rate at which a league-average shooter would convert a given shot, based on everything we can measure about that shot's circumstances. Player-tracking cameras record the conditions of every attempt, and a model uses those conditions to assign each shot an expected make probability before we know the outcome.

The tracking metrics that formalize this are often labeled with shorthand like qSQ — quantified shot quality, the expected effective field-goal percentage of the shots a player takes — and qSI — quantified shooter impact, roughly how much better or worse a player shoots than that expectation. The names vary by provider, but the concept is consistent: one number for the quality of the shots, a second number for what the shooter does relative to that quality. A shot worth an expected 55% that a player makes is right on expectation; the same shot from a player who hits it 65% of the time over a season is the mark of a shooter adding value beyond his shot selection.

What makes a shot “good”

A shot-quality model is only as good as the factors it feeds on, and the major ones are well established. The biggest drivers:

Distance from the basket. The single strongest factor. Shooting accuracy falls off reliably with distance — a shot at the rim is worth far more in expectation than a long jumper. This is the same relationship that makes the corner three more efficient than a longer above-the-break three: closer means more makes.

Defender proximity. How close the nearest defender is, and from what angle. A shot with a defender's hand in the shooter's face converts far worse than the identical shot taken in open space. Tracking measures the distance to the closest defender at the moment of release, which is the heart of the "openness" of a shot.

Openness and contest. Related to proximity but broader — whether the shot is classified as wide open, open, tight, or very tight, and whether a defender is closing out or has already arrived. A catch-and-shoot look against a scrambling defense is a very different shot from a pull-up against a set defender, even from the same spot on the floor.

Shot type and movement. Catch-and-shoot versus off-the-dribble, the number of dribbles before the shot, whether the shooter was moving or set, the time left on the shot clock. A spotting-up catch-and-shoot is easier than a pull-up off a live dribble, and the model knows it.

Feed all of this in, and the model returns an expected make probability for the shot — an estimate of how a typical shooter would fare in exactly those circumstances.

Separating shot selection from shot-making

The real payoff of shot quality is the split it enables. Once every shot carries an expected make probability, you can describe a player along two independent axes.

Shot selection is the average quality of the shots a player takes — the sum of his expected field-goal percentages. A player with high shot-selection quality is generating good looks: open shots, close shots, the efficient zones. A player with low shot-selection quality is living on difficult shots, whether by choice, by role, or because that's all the defense allows him.

Shot-making is how a player performs relative to that expectation — his actual field-goal percentage minus his expected one. A positive shot-making figure means he converts harder shots than a typical player would; a negative one means he leaves value on the table even on the looks he gets. This is the closest the data comes to isolating pure shooting talent from the situations a player finds himself in.

That distinction reframes a lot of arguments. A high-volume scorer who shoots a middling percentage might be a great shot-maker stuck with a brutal shot diet — he's making tough shots, just not enough of them to post a gaudy percentage. Conversely, an efficient role player might be a modest shot-maker who simply gets fed a steady diet of layups and wide-open corner threes. Both can be valuable; shot quality tells you which kind of valuable, and that's information no traditional stat carries.

The caveats

Shot quality is a model, not a measurement, and it inherits the limits of every model. It can only account for the factors it's fed: a model that doesn't know the shooter was off-balance, or that the defender was a foot taller, will misjudge those shots. Defender identity and height matter and aren't always captured. And like any rate stat, shot-making figures are noisy over small samples — a hot week of tough shots can make a mediocre shooter look elite until the sample grows. A player who beats his expected percentage over a handful of games may simply be running hot, the same way a single three-heavy game can mislead, a problem I cover in the three-point variance problem.

There's also the matter of provider disagreement: different tracking vendors build their expected-value models on different factors and different weightings, so two shot-quality systems can rate the same shot differently. As with any modeled metric, the number is an estimate with error bars, not a verdict — useful as a sharper lens, dangerous if treated as ground truth.

The takeaway

Shot quality is the answer to a question field-goal percentage can't touch: not did the shot go in, but how hard was it? By modeling the expected make probability of every attempt from its distance, its defender proximity, and its openness, tracking data finally lets us separate the two things a shooting percentage has always blurred together — the shots a player chooses and how well he shoots them. The next time a player's percentage looks underwhelming, ask the shot-quality question before you judge him. He might be a brilliant shot-maker drowning in a terrible shot diet — and that's a distinction worth far more than the raw percentage ever told you.