A quick note

Ads keep the free version running

It looks like an ad was hidden or removed. Advertising funds our match data and free predictions. You can allow ads for this site to support the free version.

Methodology

How AI predictions are produced on BasketPredict — inputs, models, scoring, and known limitations.

Data inputs

The available context for a prediction can include:

  • Game metadata: tip-off time, arena, home/away, league, round.
  • Recent stored results for both teams.
  • Head-to-head record between the two teams over their last meetings.
  • Home/away split — how each team performs at home and on the road.
  • The date and source of the available game context.

For background on this metric, see Elo rating.

Models

Predictions are generated on request with AI using the available match and team data. Each published analysis is a dated snapshot.

The models and amount of data can vary between games. Check the publication time and current team news before interpreting a forecast. For the underlying probability model, see probability theory.

Prediction scoring

User predictions are evaluated against the final winner:

  • Accuracy is the share of correct winner picks among graded predictions.
  • User polls do not collect exact scores. Exact-score accuracy is reported separately for eligible AI forecasts.

Calibration

A probability is an estimate, not a measured guarantee. The accuracy page reports observed outcomes for its stated sample; past accuracy does not guarantee future results. For the statistical concept, see Calibration (statistics).

Known limitations

The available context may omit injuries, rest days or last-minute lineup changes. Published forecasts can become outdated; read them alongside current team news.

About us · Editorial policy