Here is a fact that sounds like a paradox and isn't. In 2023-24, NBA players shot a worse percentage from 23 feet than from 20 feet, and the 23-footer was still the far better shot. We binned 25,000 real shots by their exact distance and watched two curves diverge: accuracy falls steeply over the first few feet and then keeps sliding, with one tiny bump, as you back up from the rim, but the value of a shot falls with it right up until the three-point line, where it jumps off a cliff in the good direction. That discontinuity is the single most important line in modern basketball, and you can see it in the raw make rates.

The smooth part: accuracy just decays

Shooting a basketball gets harder the farther you stand from the hoop. No surprise there, but it is worth seeing the shape of the decline: a cliff right off the rim, then a long, gentle slope. We bucketed every non-backcourt shot in the bundled shot file by distance and took the make rate in each band. It is almost, not quite, a clean staircase down.

Two-panel bar chart of 25,000 NBA shots from 2023-24 binned by distance. Left panel: field-goal percentage falls from 69.5% in the 0-3 foot band to 34.4% from 26-plus feet, flat around 43% from 6 to 13 feet, with no break at the three-point line. Right panel: points per shot falls from 1.39 at the rim to 0.81 on the 18-21 foot long two, then jumps 40% to 1.14 in the 22-25 foot band, where shots are worth three.
The same eight distance bands, two different stories. Accuracy (left) drops steeply off the rim, flattens through the short mid-range and keeps falling right across the arc; value (right) bottoms out at 0.81 points per shot on the long two, then leaps to 1.14 the moment the scorekeeper pays three. The dotted rule marks where shot value flips from 2 to 3. Source: bundled data_layer/nba_league_shots.csv (25,000 attempts; league shot-chart records; where the file was downloaded from was not recorded, see data_layer/SOURCE.txt), backcourt heaves excluded. Charted by our chart script with a stamped provenance footer.
DistanceAttemptsFG%Points per shotShare of all shots
0–2 ft (at the rim)5,95269.5%1.3923.9%
3–5 ft3,13446.5%0.9312.6%
6–9 ft2,24543.4%0.879.0%
10–13 ft1,71743.6%0.876.9%
14–17 ft1,29942.6%0.855.2%
18–21 ft69040.6%0.812.8%
22–25 ft6,19038.1%1.1424.8%
26+ ft3,72034.4%1.0314.9%

Source: bundled data_layer/nba_league_shots.csv, 25,000 shot-level records (league shot-chart records; where the file was downloaded from was not recorded, see data_layer/SOURCE.txt), backcourt heaves excluded. FG% and points per shot computed by binning SHOT_DISTANCE; points per shot = shot value (2 or 3) × make rate. Two- vs three-point value assigned from SHOT_TYPE.

Read the FG% column top to bottom: 69.5, 46.5, 43.4, 43.6, 42.6, 40.6, 38.1, 34.4. Almost all of the fall happens in the first few feet, 23 points between the rim band and the 3-to-5-foot band. After that the line is nearly flat for a while, and it ticks up once, from 43.4% at 6 to 9 feet to 43.6% at 10 to 13 feet: two-tenths of a point on about 2,000 shots a band, well inside sampling noise, so the honest reading is “flat”, not “better”. From 14 feet out it falls again, band after band. The two 22-to-25 and 26-plus bands, which are three-point range, keep falling: a 26-footer is harder than a 23-footer and the make rate says so. Distance is not the only thing in a make rate (who takes the shot and how closely it is guarded vary by band too), but in this file nothing about crossing the arc changes the downward slope. If points were all that mattered, every shot would be a layup.

The cliff: value doesn't

Now read the points-per-shot column instead, and the smooth staircase breaks. Value falls in lockstep with accuracy through the entire two-point range: 1.39 at the rim, down through 0.93, 0.87, 0.85, bottoming at 0.81 in the 18-to-21-foot band, the long two, the worst-paying real estate on the floor. And then, at exactly the point where the accuracy staircase keeps descending, the value line leaps: the 22-to-25-foot band pays 1.14 despite a make rate of only 38.1%.

That is the cliff. A shot from 20 feet returns 0.81 points; a shot from 23 feet returns 1.14, a 40% jump in value for backing up three feet and shooting a lower percentage. Nothing about the shooter's skill changed across those three feet. The only thing that changed is that the scorekeeper started awarding three points instead of two. The entire modern reshaping of NBA offense is a response to this one discontinuity, and it is why we keep coming back to the arithmetic of points per shot, set out below.

Put the two shots side by side and do the multiplication by hand, because the whole lesson fits in two lines. Points per shot is nothing more than the point value times how often the shot goes in.

The long two, from the data: a two-point shot hitting 40.6% of the time is worth 2 × 0.406 = 0.81 points per attempt. The above-the-break three: a three-point shot hitting 38.1% is worth 3 × 0.381 = 1.14 points per attempt. The three-pointer is made two-and-a-half percentage points less often, and it still pays 40% more. A player who shoots 41% on long twos and 38% on threes should never take the long two; the extra point swamps the accuracy gap and then some. Multiply that edge across a season of attempts and you have the reason the 18-to-21-foot band is down to 2.8% of all shots while the 22-to-25 band is up at 24.8%. Teams didn't abandon the long two because they forgot how to shoot it. They abandoned it because 0.81 sits below 1.14, and there is a bright painted line that makes the difference legal.

The formula, and a tool to try it

Points per shotPPS = point value × FG%

That is the whole calculation, and it is why a barely-better-than-one-in-three three beats a coin-flip two: 3 × 0.36 = 1.08 against 2 × 0.50 = 1.00. The explorer below lets you set how a team spreads its attempts across four zones (rim, mid-range, corner three, above-the-break three) and how well it shoots each, then computes points per shot by zone and expected points per 100 shots. Put the long two at this file’s 40.6% and it pays 0.81; no realistic make rate from 18 feet catches an ordinary three.

Shot-Value Explorer: points per shot = point value × FG%. A 42% mid-range two yields 0.84 points/shot; a 39% corner three yields 1.17. Enable JavaScript to use the interactive tool, or try the calculators page.

Points per shot is a per-attempt number, not a per-possession one: it ignores turnovers and offensive rebounds, and a possession has to produce a clean look before any shot value applies. How possessions are counted picks up that thread.

Why the rim still wins

The cliff at the arc gets the attention, but the tallest number in the table is at the other end: 1.39 points per shot from inside three feet, on a 69.5% make rate. The rim wins not because the reward is large (it is only two points) but because the success rate is enormous, and no amount of three-point value catches a shot that goes in seven times out of ten. This is why the modern floor has two peaks and a starved middle, the exact distribution we mapped in where the NBA actually shoots from. The layup and the open three are the whole menu; everything between 4 and 21 feet is the food nobody orders. The corner three, which sits a foot closer than the top of the arc, is just the sweetest version of the same trade.

Honest limitations

Distance is not difficulty. The bins treat every 18-footer as the same shot, but a wide-open catch-and-shoot 18-footer and a fadeaway 18-footer over a contest are worlds apart. FG% by distance folds shot quality, defender proximity, and shot type into one average. That is exactly the gap that shot-quality tracking exists to fill, and it means the value cliff describes the average shot, not any particular one.

The metric ignores fouls entirely. Points per shot counts only attempts that end in a field-goal try. A drive to the rim that draws a shooting foul scores zero in this table even though it might net a point and a half from the line. That undercounts rim pressure specifically, and it is the reason true shooting percentage folds free throws back in. The rim's 1.39 is, if anything, an understatement of its real value.

Selection runs both ways at the rim. That 69.5% make rate is partly a product of shot selection, not just proximity: the rim attempts that survive into the data are the ones players chose to take after passing up the contested drives. Read it as the payoff of good rim shots, not proof that any drive is a good idea.

Late-clock situations flip the math. When a possession is dying and the defense has taken away the rim and the arc, the devalued mid-range jumper is suddenly the correct read, because a 0.81 shot beats a 0.0 turnover. That is the whole nuance of the mid-range's evolution: the shot is inefficient on average and indispensable in context.

Accuracy falls smoothly; value jumps at the arc

Two curves, one floor. Accuracy falls smoothly and honestly with every foot you back away from the rim, and it never lies: a 23-footer really is harder than a 20-footer. But value does not follow accuracy across the three-point line; it falls off a cliff in the good direction, because the rulebook pays a full extra point for a shot that is only a little harder. Run your own numbers on the Shot-Value Explorer and you can watch it happen: crank the long-two make rate up to something no one shoots and it still can't catch an ordinary three. The modern NBA is just a league that read this table and started standing on the far side of the line.

Sources & Further Reading

  • Shot-level data: bundled data_layer/nba_league_shots.csv (25,000 attempts; league shot-chart records; where the file was downloaded from was not recorded, see data_layer/SOURCE.txt). FG% and points per shot by distance computed with a pandas cut on SHOT_DISTANCE.
  • The geography of shot selection and the death of the long two: Kirk Goldsberry, Sprawlball.
  • Expected-value framing of shooting: Dean Oliver, Basketball on Paper.
  • Shot-zone make rates and stat definitions: Basketball-Reference Glossary; live shot data at NBA.com/stats.
  • Three-point line geometry and court dimensions: official NBA Rulebook.

How this was made

Published by NBAAnalytic from public data and the sources named in this piece. How this site is made · the methodology