Next-24h RTP forecast per game: observed RTP blended with the configured value by sample size, with 95% intervals from real tracked bets.
The model blends what we have actually observed for a game against its configured RTP, weighted by sample size: the more bets we have recorded in the last seven days, the more the observed figure dominates. With little data the forecast sits close to the theoretical value, which is the honest default.
Confidence intervals come from the dispersion of hourly RTP within the window, so a game whose hourly results swing wildly gets a wide band and a stable one gets a narrow band. Games with fewer than thirty bets in the window are excluded entirely.
A game is flagged hot or cold when its predicted figure deviates by more than a percentage point from its configured RTP. That deviation is information about measurement — either the sample is small, or the build differs from what we expect, or the recent run has been unusual.
It is not information about your next spin. Rounds are independent; a cold game is not owed a payout and a hot one is not on a run you can join. Anyone selling the opposite is selling something.
Persistent deviation at high confidence over thousands of bets is the interesting case, because the most common explanation is not luck — it is that the operator is serving a different RTP build than the one we hold as theoretical. That is measurable, and we track it per casino in the RTP builds report.
No. Independent rounds have no memory, so no timing strategy derives from these numbers. The forecast describes measurement, not opportunity.
Small samples deviate naturally, and some operators serve lower-RTP builds of the same title. Confidence and volume tell you which explanation is more plausible.