NBA · 2026-27 · Week 2

The Shot Diet

One week in, a team's 3-point percentage is mostly noise. Where it shoots from is not. Five seasons of games show why, and which teams have really changed.

By Shawn Nafziger · Data through Oct 27, 2026

0.69How well a team's after 4 games predicts the rest of its season (, five-season average)

0.09How well a team's 3-point percentage after 4 games predicts the rest of its season: essentially not at all

3.5 ppHow far a team's 3-point attempt rate normally wobbles in its first 4 games. Smaller moves don't mean much

How to read these: the first two numbers are correlations, a score from 0 to 1.0 for how well a team's first 4 games predict the rest of its season. 0 means they tell you nothing; 1.0 means they predict it perfectly. Around 0.7 and up, early numbers are a good guide.

Choices settle fast. Makes don't.

Whether to shoot a three is a decision, made by coaches and players before the ball goes up. Whether it goes in is partly luck. Over the last five seasons, a team's 3-point attempt rate after just 4 games lined up with the rest of its season at 0.69, and between 0.62 and 0.73 in every one of those seasons. Its 3-point percentage lined up at 0.09. Even after 20 games it only reaches 0.39.

So when Chicago opens the season making 23.8% of its threes, down from 35.6% last season, history says to wait. When a team changes how often it shoots threes, pay attention.

How well a team's first games predict the rest of its season

Correlation across the 30 teams, average of five seasons (faint lines: each season)

3-point attempt rate3-point percentage

Games 1–k vs games k+1–82, regular seasons 2021-22 to 2025-26, from team totals. Source: NBA Stats API.

Who changed their shot diet

Across the league, threes are 42.2% of shots so far, close to last season's 41.5%. A few teams moved much further. The chart compares each team's share of shots that are threes with its 2025-26 season. The blends this season's games with last season's, as in The Early-Season Mirage.

Two teams have changed clearly enough to rule out a noisy week: Detroit (+8.2 so far, best estimate +5.7) and Dallas (+6.2 so far, best estimate +4.5). Orlando (+8.4), the Lakers (+7.2) and Oklahoma City (−7.4, the other way) look like real shifts too, but their still reach zero.

Change in 3-point attempt rate vs 2025-26, in percentage points

Regular season through Oct 27 (56 games). Hover a team for details.

So farBest estimateClearly changed90% range

Show as a table

Team	Games	3PA rate 2025-26	So far	Change	Best estimate	90% range	3P% so far

Where the shots moved

Each team got there differently. Pick a team to see its and who is taking the threes.

Detroit traded rim attempts (−12.2 pp) for threes above the break (+5.9). Duncan Robinson is taking 9.5 threes a game, up from 7.0.
Dallas has Kyrie Irving back. He didn't play in 2025-26 and is taking 8.3 threes a game. Above-break threes are +7.7 pp, rim attempts −6.7.
Orlando: Desmond Bane (5.2 to 9.3 a game) and Tristan da Silva (4.2 to 8.7) have roughly doubled their threes.
The Lakers moved shots out of the paint (−5.8 pp) to above the break (+5.3).
Oklahoma City went the other way: more at the rim (+5.5), fewer above-break threes (−6.3), even though Shai Gilgeous-Alexander is taking 7.0 threes a game, up from 4.4.

Share of shots by zone

Thin bar: 2025-26. Thick bar: 2026-27 so far. Right: change in percentage points. Below: who is taking the threes.

Player	Games	3PA per game	2025-26	Change

3-point attempts per game played, from box scores. 2025-26 counts every team the player played for. Source: NBA Stats API.

These are choices, so they can change again with an injury, a new lineup or a trade. But they are signals, not noise, and that's rare in week 1.

Method

Data: team game logs for every regular-season game from 2021-22 to 2025-26, and 2026-27 through Oct 27 (56 games). Shot locations: 219,131 shots from 2025-26 and 9,993 from 2026-27, from the NBA Stats API via the open-source nba_api client.
3-point attempt rate = 3PA ÷ FGA, from season totals (free throws aren't included). 3P% = 3PM ÷ 3PA, also from totals.
Predictive test: Pearson correlation across the 30 teams between games 1–k and games k+1–82, averaged over five seasons, so early games never appear on both sides.
Wobble: the standard deviation of (3PA rate over a team's first 4 games − its full-season rate), pooled over 150 team-seasons: 3.5 pp. For scale, teams' rates change by a typical 3.7 pp from one season to the next.
Best estimate: a linear regression of rest-of-season 3PA rate on (a) the rate so far and (b) last season's rate, fitted at the same game count over four season pairs (120 team-seasons). The 90% range comes from the regression's typical error. "Clearly changed" means the range doesn't include zero.
Zones use the NBA's shot-zone labels: Rim = restricted area; Paint = the rest of the lane; Midrange = other 2s; corner and above-break 3s. Backcourt heaves are left out. Zone shares compare one season with a few games, so small zone changes are noise too.