How we test: the fair wheel, the simulations and what we publish

How the results are made

Every spin in the app is drawn from your browser's cryptographically secure random generator, the one it uses for security, never from the everyday random function. A draw that would make some pockets more likely than others is thrown away and drawn again, so each of the 37 pockets, 0 to 36, is exactly as likely as any other.

The generator sits in a part of the app that reads nothing from the rest of it. It cannot see your bets, your balance or the strategy you are running, so no result can react to them. Winning bets are paid from the standard European table that the payout chart shows: 35 to 1 for a single number down to 1 to 1 for red or black. The Fair results panel in the app explains the same in a few lines.

How the strategy results are built

The strategy pages publish what 10 classic betting systems and our own system did over 100,000 simulated sessions each, 1,100,000 sessions in all. The classic systems all play the same kind of session:

• Every classic session starts with 500 points and a base bet of 5 points, the amounts the Strategy Simulator starts with, and plays up to 100 rounds, or until the points run out.
• Every round keeps to the app's standard table limits: inside bets from 1 to 100 points per spot, outside bets from 5 to 1,000, and at most 2,000 points on the table. When a system calls for a bet the table refuses, it stops there, as it would in the app.
• The systems run on the app's own code: the same strategy rules and the same backtester the Strategy Simulator uses, read without change.
• The spins come from a seeded generator over the 37 pockets of a fair single zero wheel, and every classic system is dealt the same sequence of spins. That is the one place a seeded generator is used instead of the browser's: it makes the whole run repeatable (run again, it gives the same numbers), and it makes the comparison between systems fair. The seed only fixes the order of the spins: every pocket is as likely as on a fair wheel, to within one part in a hundred million.
• The dataset was generated on 2026-09-23.

Our own system plays its own kind of session, so it is measured the way the app's leaderboard measures it, over 100,000 sessions on the same kind of seeded fair wheel, at the leaderboard's fixed bankroll.

What we publish, and what we keep private

For each classic system we publish its rules, how often a session finished ahead or ran out of points, the average, median and worst result, and the whole spread of results in charts. Pages about the game itself, such as the payout chart, are exact math from the same payout table the app pays with, and say so.

For our own system we publish the outcomes only. Its rules, how it sizes its bets and how it picks its numbers stay private. Nothing kept private changes the conclusion: no betting system beats the house edge, ours included.

Who pays for this

Nobody but us. The app is free, carries no advertising and sells nothing: there is nothing to buy, win or cash out. We take no money from casinos or betting sites, carry no casino advertising and link to no casino. The only sites about gambling we link to are services that help people whose gambling has stopped being a game.

The Roulette Assistant is operated by Galya OÜ.

How to reach us

Found a figure that looks wrong, or want to know how a number was made? Write to support@therouletteassistant.com. It reaches a person, and we answer in English.

Questions about the app itself are answered on the support page. Writing about the app? The press kit has short descriptions, key facts, the logo and screenshots.