Trust Center

Model Cards & Calibration

Every model that backs a shown prediction publishes a card with its known strengths, known weaknesses, and grading-based calibration. Calibration is the model's measured error on graded games — never a marketing win-rate.

Public

Every pick is shown — none are hidden.

The sample sizes below count every graded prediction a model has made, wins and losses alike. We do not drop, cherry-pick, or reset results to flatter a number. Calibration is the mean absolute error between what a model predicted and what actually happened on graded games — a measure of how close it tends to be, not a claim about any future outcome.

The Trust Center is unavailable right now: model_card_calibration read failed: TypeError: fetch failed