Three forecasts.
Three different questions.

First points, not just the first field goal

Our main target is the player credited with the game's first points. A made free throw counts; a missed free throw does not. A made two-pointer or three-pointer also counts. If a free throw comes before the first field goal, the two categories can have different winners. Our definitions do not determine any sportsbook's settlement rules.

What goes into a forecast?

Our statistical models consider historical player performance, confirmed starters, team matchups and shooting tendencies. They use earlier games, not the game being predicted or future outcomes. LLMs assist development and research; they do not directly generate the displayed probabilities.

Forecasts can be unavailable when lineups or supporting data are incomplete. Missing information is not presented as a confident prediction.

1. First possession

The first card estimates which team is more likely to gain opening control, informed by previous opening possessions and the expected matchup. Getting the ball first does not guarantee taking the first shot or scoring first. We grade this forecast only when normal opening control can be verified.

2. First-points player

The second card highlights the highest estimated first-points probability among the ten confirmed starters. The ranking considers player scoring history and team context. Those ten probabilities total 100%; a leading player is still more likely to miss than score first when their probability is below 50%. Tied leaders are not forced into a pick.

3. The selected player's method

The third card estimates the scoring method if our selected player scores first, informed by their earlier scoring and shooting history. Methods include free throws, three-pointers and different types of two-point baskets. Its percentage is conditional: it is not the chance of getting both the player and method right.

How we measure performance

“Accuracy to date” means correct decisive picks divided by graded decisive picks. Today's games are excluded using Eastern Time. Ties and unresolved evidence are excluded and reported separately, so each card can have a different sample size.

Method accuracy counts only games where our selected player actually scores first and the method is known. A wrong-player result is a combined player-and-method miss, not a conditional-method trial. Non-starter first scorers count as player-pick misses but are excluded from probability-quality scores requiring the ten-player choice set.

Our research also checks how well the probabilities match observed outcomes, not just how often the top pick wins. Accuracy alone does not establish profitability.

Historical replay versus a saved pregame forecast

Historical replays are reconstructed after games, using earlier history. They may include corrected data and lineups whose pre-tip availability is unknown. Saving a replay for faster loading does not make it a pregame prediction.

Saved pregame forecasts are recorded before tip-off and remain fixed while results are graded separately. We distinguish these sources in our reports. Historical research is not proof of future performance.

What these estimates cannot promise

Lineup changes, limited history, data errors and unpredictable play all matter. This is an independent research and entertainment project, not an NBA-endorsed service. No outcome or financial return is guaranteed.

Explore past predictions and results and Model Research for the current evidence. For questions or a suspected data issue, use our Contact page.