What the model has learned

Loading…

Every night at 4:00 AM CT, after the engine grades the slate, TopCheddy re-reads every graded game of the season and asks two questions: how much should the card trust the engine's probabilities versus the betting market, and which situations have actually paid? The answers reshape the next day's card. The engine's own numbers never change; the card's reading of them does.

Last night's findings

Loading…

Trust in the model, by market

Loading…

Record under the learned model

Loading…

Teams the model keeps getting wrong

Loading…

Situations — what has and hasn't paid

Loading…

Calibration

Loading…

Training runs

Loading…

Personal project. For entertainment and informational purposes only; not betting advice.