Shot-making over expected (xeFG)

Effective field goal percentage is a genuine improvement on raw FG% because it credits a three-pointer with the extra half-point it is worth. What it still cannot do is tell you why a number is high. A 65% eFG% can mean a player is an extraordinary shooter, or it can mean he only shoots from three feet, or some blend of both. The stat pools skill and role into one figure and hands it to you unlabeled.

That matters because the two ingredients behave completely differently. Shot selection is largely a function of role, scheme, and who else is on the floor. A center in a modern pick-and-roll offense does not choose to shoot dunks and lobs so much as the system chooses for him. Shot making is closer to a personal skill, the part that travels when a player changes teams or gets handed a bigger workload.

Expected eFG% is how you separate them. You take every league-average efficiency figure for each area of the court, weight those averages by how often the player actually shoots from each area, and you get the eFG% an anonymous league-average player would have produced on that exact shot diet. That is the shot-selection half. Subtract it from what the player really shot and what is left is the shot-making half.

This is the single most common misreading in basketball shooting stats, and it is the reason the metric exists. Shots at the rim go in at a far higher rate than anything else on the floor, which is exactly why teams work so hard to generate them. A player whose entire shot profile is dunks, lobs, and putbacks will therefore sit near the top of any eFG% leaderboard without ever having made a difficult shot.

Run that player through an expected model and the leaderboard rearranges. His expected eFG% is also enormous, because a league-average player taking only rim attempts would also convert at a high clip. His shot-making over expected can land near zero, or below it, even while his raw eFG% is elite. Nothing about that is an insult. Finishing at the rim at league rate while occupying two defenders is valuable basketball. It is just not the same skill as making shots that are hard to make.

The same logic runs the other way. A high-volume guard who lives on pull-up threes and long twos will show a mediocre raw eFG% and a low expected eFG%, because his shot diet is genuinely harder. If he beats that expectation, the gap is telling you something real about his shot making that his percentages hid.

We build expected eFG% from shot location. For each of the court zones the NBA publishes (restricted area, paint outside the restricted area, mid-range, left corner three, right corner three, and above-the-break three) we compute the league-wide efficiency, weight those figures by the player's own attempt share in each zone, and convert to an eFG% basis so that threes carry their extra credit. Subtracting that expectation from the player's real eFG% gives the number we publish, in eFG percentage points.

It is a deliberately simple model, and simple has an advantage: every input is a shot count you can check, and there is no black box between the raw data and the answer. The full technical writeup, along with every other metric we compute ourselves, is on our methodology page.

Being honest about the boundary is more useful than pretending there is not one. Our expectation is built from where a shot was taken and nothing else. It does not know any of the following.

Several public models do more than we do here, and they are worth reading. The NBA's own expected field goal percentage, explained at nba.com/news/intro-to-expected-field-goal-percentage, uses optical tracking to model defender contest posture, shooter orientation, and balance, not just court location. BBall-Index publishes Shot Making and Overall Shooting Talent, which build an expected eFG% from shot difficulty rather than location alone (openness, location, shot type) and add a weighting for whether the player created the attempt himself, and they report that their shot-making measures hold up far better year over year than raw eFG% does. Both are strictly richer inputs than ours.

We do not carry that tracking data, and we would rather ship a location model we can fully describe than approximate a contest model we cannot verify. The same principle applies elsewhere on the site: we publish Ben Taylor's Box Creation because it runs on box-score inputs we hold, and we deliberately do not approximate his Passer Rating because it needs tracking we do not have.

The practical consequence is that our figure should be read as a first-pass separation of selection from making, not as a finished shot-difficulty model. Where it disagrees with a tracking-based version, the tracking version is more likely to be right about why.

Treat it as a modifier on efficiency, not a ranking of players. The useful reading is comparative: two players with the same eFG% but different signs on shot-making over expected are doing genuinely different jobs, and the one beating expectation is the better bet to keep his efficiency if his role or supporting cast changes.

Volume still gates everything. A small number of attempts produces a large gap by chance alone, so a big figure on a thin sample is noise until it repeats. Look for players who clear expectation across multiple seasons rather than one.

Finally, do not treat it as an impact metric. It says nothing about defense, playmaking, turnovers, or how many possessions a player shoulders. For an all-in-one estimate of on-court value, our IPM page is the right stop, and for shot creation, box creation. This page answers one narrow question well: given the shots he takes, does the ball go in more often than it should?