A walkthrough of the methodology — from spin entry to statistical rankings and distribution analysis.
Open a session and tap each roulette outcome as it lands. The app records the number, timestamp, and running session context. Spin entry is designed to be fast — a full-screen number pad with haptic confirmation so you can keep pace with the dealer.
You can track multiple sessions simultaneously (multiple tables at the same casino, for example). Each session maintains its own independent history.
After each spin, statistical methods recalculate in the background. Some are descriptive diagnostics; formal evidence gates control which estimators may contribute to scored rankings.
On a fair European wheel, each of the 37 pockets has the same long-run probability. Hot/cold frequency, Z-score, and related methods summarize how the recorded counts differ from that uniform baseline. A short-run deviation alone is not evidence of bias.
Neighbour clustering and sector methods map each outcome to its physical position on the wheel and summarize whether recorded results concentrate in contiguous regions. Spatial clustering can be investigated, but it is not by itself proof of a mechanical cause.
Markov chain analysis looks at transitions in chronological data. Because a roulette transition table is sparse, the method contributes only when formal independence testing finds supported serial dependence; otherwise it receives no scored weight.
Sample size and repeated checking matter. Roulette Ace uses evidence gates designed for ongoing monitoring so unsupported deviations stay exploratory instead of being treated as scored findings. Insufficient evidence keeps estimates close to the uniform baseline.
Compact ranking rows can appear after at least 25 recorded spins, but each method has its own evidence requirements. No fixed spin count guarantees that an estimator will become eligible for scored output.
Bias Estimator A and Bias Estimator B estimate the distribution represented in the recorded sample. They are separate from the statistical engines and do not forecast an independent next spin.
Both are trained instances of the same bundled native estimator architecture. Their regularized distribution estimates support comparison and ranking overlap; displayed rank or signal scores are not next-spin probabilities.
Inference runs on your phone. Your account-owned tracking records separately synchronize through Firebase when a connection is available.
Engine Picks shows statistical rankings, while the native-estimator row remains separate. Consensus shows overlap between eligible statistical and estimator rankings. Rank and signal scores summarize the recorded sample; they are not next-spin probabilities.
The Table Quality score summarizes irregularity in the recorded sample. A higher score means the sample departs more from the uniform baseline; a lower score means it remains closer to that baseline. The score is descriptive, not proof of a mechanical cause or a next-spin forecast.
The Wheel Bias screen visualizes the physical wheel and highlights arcs with the strongest recorded deviation. Spatial concentration is a diagnostic worth investigating, not proof of mechanical bias; a scattered pattern can be consistent with ordinary random variation.
Roulette Ace cannot predict the next number on a fair, well-maintained wheel. A genuinely random roulette outcome has no statistical relationship to previous spins — that is the definition of randomness.
What it can do is detect when a wheel is not consistent with a uniform baseline — and summarize the strength of any recorded deviation while keeping unsupported results exploratory. This cannot by itself establish a physical cause. Roulette Ace is also an entertainment tool: tracking and analysing your sessions can be interesting even when the results are compatible with ordinary random variation.
Roulette Ace is for entertainment and educational use only. No real-money gambling. No guaranteed outcomes. 18+