Identify the type of number
A historical statistic summarizes recorded games in the selected sample. A projection estimates production in the chosen market’s units. A model probability estimates a chance under model assumptions. A research rating is an ordering measure unless the view explicitly defines it otherwise.
Implied probability comes from a quoted price. A no-vig estimate adjusts market prices for margin. Neither label by itself means the app measured a model’s accuracy.
Open the supporting detail
- Check the sample count, date range, and any unavailable or stale markers.
- Open the game log or source receipts offered by the view.
- Use Model & evidence or Data & methodology to inspect the actual method and limitations.
- Enable Dev mode when you need the additional inputs or formulas exposed by the research page.
Treat simulation output as conditional
A simulation uses the inputs and assumptions selected for that run. Increasing the number of runs can reduce simulation noise, but it does not prove that the model represents the real event accurately.
A Monte Carlo interval is not the same as a validated prediction interval. Similarly, a difference in probability points is a comparison between estimates, not a guaranteed advantage. Keep the source context attached when saving or discussing a result.