Shot lab. Choosing a release beats chasing green.

Free Throw Challenge asks you to read a moving rim. The green prompt is tempting, but it covers a region of the court, and the hoop can cross that region in either direction. We compared five release policies against the game’s shooting model to find which part of that information actually helps.

Study checked November 12, 2026. Each policy uses 100,000 isolated shot trials with a fixed random seed. The results measure the shooting model under the conditions described below.

The experiment

The ball locks onto the rim’s position when you release. It keeps that trajectory while the hoop travels on. The experiment advances the hoop first and the ball second on each step, matching the game’s collision order. A ball within 16.8 pixels of the rim center is a clean make. A wider contact, below 34 pixels, has a 30% chance to drop.

For each trial, the hoop’s starting speed is sampled uniformly from the game’s 0.015–0.025 range. Every turnaround draws a new speed. We test a release at a prescribed location or sample a location within the policy’s range. The 24-pixel point is approximately halfway between the center and the end of the rim’s roughly 47-pixel travel.

Modeled make rates for five release policies
Release policyMake rate
Unchosen moment23.7%
Anywhere in green, either direction24.7%
24 pixels out, either direction48.9%
24 pixels out, moving outward98.3%
24 pixels out, moving inward0.0%

Direction carries most of the answer

Under the study’s uniform-angle sampling, releasing somewhere within green barely improves the modeled make rate over releasing somewhere in the full swing. Green narrows the location but leaves the direction unresolved. The hoop can be rushing away from the ball’s eventual target or turning back toward it.

Always releasing 24 pixels from center, with no direction check, raises the rate to about half. Waiting for the outward-moving crossing produces the strongest result in this study. At that point the rim travels to the edge, reverses, and comes back toward the spot where you aimed.

The inward policy recorded zero expected makes in these sampled trials. That result applies to the tested 24-pixel position and speed range. Other release points have their own probabilities, as the manual’s timing table shows.

Why streaks magnify a timing improvement

The third consecutive make and each one after it earn three points. The first two makes of a streak earn two. A better make rate therefore raises both the number of baskets and the chance that a basket receives the extra point.

For 100 independent attempts with make probability p, the expected score is 200p + 98p³. Every attempt can earn the base two points. Attempts three through 100 can also earn one bonus point when that attempt and the preceding two all go in.

Theoretical points over 100 independent attempts, with the streak rule
PolicyExpected points
Unchosen moment48.8
24 pixels out, either direction109.2
24 pixels out, moving outward289.6

This calculation isolates the scoring benefit of consistency at a fixed number of attempts. A sixty-second round limits how many attempts you can take, and waiting for a specific crossing costs time. A policy with a higher make rate can still produce fewer total attempts. Your final score depends on both factors.

A practice routine you can use

Start in timing practice. Watch a full swing without shooting. Identify the center, the two endpoints, and the point halfway out on each side. Then take one shot per outward crossing. Skip the crossing if you notice it late, and look for the next opportunity on the other side.

When you can recognize direction consistently, try the weekly challenge. Compare accuracy and score across your own rounds. A higher score with worse accuracy may simply mean you took more attempts. A steady accuracy improvement gives you a more repeatable base for building streaks.

Method, downloads and limits

The experiment imports the game’s actual hoop geometry. Its launch, turnaround and collision rules mirror the implementation, following the existing manual’s isolated-shot study. The downloadable script contains that model and runs with Node.js without installing packages. Seed 20261112 makes the output repeatable.

Download the reproducible shot experiment or download the measured results and source fingerprints.

For the broad release policies, angle and direction are sampled independently to define a consistent comparison distribution. Actual players choose their own timing. The model omits input delay, display behavior, attempt frequency and the buzzer. Rim contacts contribute their 30% expected success probability, which reduces sampling noise. Small differences between rerun studies can reflect the chosen seed and sampling method.

The complete controls, scoring and release-point table are in the basketball manual. Find the rest of the analysis in the guides library.