About the model

NASCARPredict estimates the probability that an entrant wins a given event. It is currently built from public NASCAR Cup results: each completed race is expanded into stable-orientation pairwise finish-order comparisons, then those comparisons update an Elo rating. A richer uncertainty model or feature layer can be added later, but it is not claimed in the live metrics today.

Field simulation

NASCAR events are large fields, not pairwise matches. We rate each entrant, add independent single-event noise, and Monte-Carlo the whole field thousands of times — reading off how often each entrant wins or finishes top-5 / top-10. More noise (the event's luck factor) flattens the favourites.

Current limitations

The site does not yet ingest official next-race entry lists or Cup standings benchmarks. The forecast page transparently reuses the latest completed Cup field as a current-field proxy, and the performance page reports only chronological walk-forward model metrics.

These are model probabilities for entertainment and analysis, not betting advice.