A finite probability model
Let K be the number of winning rounds among n independent completed rounds. K follows a binomial distribution with win probability 18/37 for the single-zero example or 18/38 for the double-zero example. The probability of k wins is C(n,k) × pk × (1 − p)n−k.
The ending net result is stake × (2K − n). The model evaluates each possible number of wins, rather than generating a random sample. The calculations use floating-point arithmetic.
Read the results correctly
Finishing positive means strictly above zero. Breaking even does not count. The 5th and 95th percentiles are the first possible outcomes whose cumulative probability reaches those thresholds, so the inclusive band contains at least 90% of outcomes.
The ending-loss threshold is not a stop-loss or ruin calculation. All n rounds are completed regardless of the intermediate balance. Reaching a loss during play and ending with that loss are different events. A zero threshold includes a break-even ending.
What this excludes
No changing stakes, bankroll constraints, la partage, en prison, multiplier payouts or side bets are included. The wheel is assumed fair and rounds independent. An actual table requires its own rules. A run of past colours does not change this model's next-round probability.
Sources & scope
Sources checked 30 September 2026. Published features and rules may change. Check the exact service, table and jurisdiction before relying on a detail.