About the engine

What it can help with

Lineup and possession-level decision support: compare who plays, shot profiles and the range of plausible scores.

What has been measured

The historical walk-forward run picked 69.4% of winners across 5,099 games. A plain rating picked 69.8% on those same games. This does not demonstrate superior winner picking. The run uses realized players and minutes.

Backtest status: stale. Historical result has no recorded engine provenance; current engine is unverified. Backtest season: 2025-26.

Uncertainty

Ranked newcomers use recruiting inputs. Unranked newcomers use position and program tier. Projected rotations receive visible high, medium or low data confidence on the simulator. Coaching-system effects remain experimental pending prospective validation.

ESPN public box scores, official athletics rosters and Bart Torvik supply the public data. Detailed methodology and evaluation tables load with the interactive page.