In the NBL, good shots are table stakes

The midrange two has halved since 2015-16 and every team now takes similar shots. What separates teams is whether the shots go in, and that is the part that does not last.

01 · The price of a shot

Rim shots and threes pay; the midrange does not

A model of how often a shot from each spot goes in, built on 223,164 shots, puts a price on every attempt. A shot at the rim is worth 1.25 points on average, a corner three 1.09, an above-the-break three 1.03. A midrange two is worth only 0.78: taking one instead of a three gives away about 24 points per 100 shots. Make rate falls off a cliff within two metres: shots inside 1.5 m go in 62.5% of the time, and a longer two goes in only 39.2%, barely better than a three (34.2%) that is worth half as much again.

The model holds up on seasons it never saw. On 72,924 shots from 2023-24 onward that it had not seen, it beats both a simple two-or-three split (log loss 0.6722) and a five-zone lookup (0.6601), at 0.6565.

02 · The quiet revolution

The NBL already shoots like a modern league

Midrange twos were 19.5% of NBL shots in 2015-2016 and 9.9% in 2025-2026. The shots moved to the rim (25.0% to 30.0%) and the three-point line (34.3% to 39.8%). Priced at the same make rates, the average shot was worth 1.007 points in 2015-2016 and 1.032 in 2025-2026, about 2.5 extra points per 100 shots from selection alone. Most of that happened by 2019-20; the easy gains have been taken.

03 · Taking and making

Making decides seasons, and it does not last

With every team taking similar shots, the difference between the top and bottom of the ladder is mostly whether they go in. Net of what they allow, teams' shot-making varies by 5.5 points per 100 shots (one standard deviation) against 1.9 for shot quality, 2.9 times as much. Shot-making tracks a team's winning margin within a season far more closely: a correlation of 0.69, against 0.28 for shot quality. In 2025-2026, the Sydney Kings (71.9% wins) were +9.2 per 100 shots on making and +0.8 on quality; the Brisbane Bullets (18.2%) were −11.2 and −1.8.

The twist is what carries over. Shot quality is a habit: a team's quality in odd games predicts its even games (0.80) and carries into next season (0.64). Shot-making is half as consistent within a season (0.48) and barely carries over (0.12). Defences choose the shot, not the result: the quality of shots a defence allows is its most stable trait (0.87 within a season), while whether opponents then make them is mostly out of its hands (0.25). A team whose opponents shoot hot is usually unlucky, not bad. For the rest of a season, making still tells you more: over the first half it predicts second-half margin (0.27, range 0.12 to 0.43) better than quality does (0.13, range −0.07 to 0.29). It is next season that it fades.

04 · Players

Shot-making is partly a skill

A player's shot-making correlates 0.40 from one season to the next (151 player pairs), well below their shot selection (0.79) but well above zero. After allowing for luck, the real spread between players is about 5.7 points per 100 shots either way. 6 of 52 players are clearly better than average (Tyrell Harrison, Jack McVeigh, Isaac Humphries, Ian Clark, Xavier Cooks, Kendric Davis) and 1 is clearly below (Mitch Norton); the rest cannot be separated from average with confidence.

Read the names with care: the big men near the top are partly a blind spot of the model, which knows where a shot was taken but not whether it was a dunk. The practical lesson for recruiting is to buy players whose shot-making has lasted across several seasons, because one hot year tells you little.

What this does not show

• Free throws are not priced. They have no court location, so shots that draw fouls are under-rewarded in the model, mostly at the rim.
• Location only. Shot type is recorded only on makes in the first two seasons, so using it would leak the outcome. The model prices where a shot was taken, not whether it was a dunk or a contested jumper.
• Player ranges overlap. Only 7 of 52 players are clearly different from average; the order within the group is tentative.
• Team-seasons are few. 100 team-seasons is enough to see the pattern, but individual club estimates are noisy and the next-season figures also absorb roster change.