Run the math.

What the Lab actually simulates

Every run above is American double-zero roulette, even-money bets: 18 winning pockets out of 38, so p(win) = 18/38 per spin and the house keeps 2/38 — the fixed 5.26% edge in the last stat tile. The table max is 100× your base bet, which is the detail that quietly kills every doubling system. The random numbers are seeded (mulberry32), which means a run is reproducible forever: same seed, same spins, same outcome — post a seed and anyone can verify your result. One highlighted run plays out on the chart, 10 shadow runs show the spread, and 200 hidden runs compute the bust-risk tile, so the verdict rests on 211 sessions rather than one lucky line.

The five systems, honestly described

Flat betting — the same stake every spin. The control group. It loses your stake × 5.26% per spin on average, and nothing else in this list improves on that number; the others just rearrange when the losing happens.

Martingale — double after every loss so the next win recovers everything plus one unit. It converts a high chance of small wins into a small chance of catastrophe: nine straight losses on even money happens roughly once every 490 attempts, and long before your bankroll runs out, the 100× table max closes the trapdoor. The bust-risk tile is where Martingale's story is told.

D'Alembert — add one unit after a loss, subtract one after a win. Gentler than Martingale, slower to bust, same expected value. Its seductive feature is that a balanced number of wins and losses shows a profit — but wins and losses don't come balanced at p = 18/38, which is the entire point.

Labouchère — write a line of numbers, bet the sum of the ends, cross them off on a win, append your loss on a loss. It feels like bookkeeping, and it books exactly the same 5.26% loss per dollar staked as everything else, with paperwork.

Reverse Labouchère — the same line run in the opposite direction: press wins, cut losses. It produces many small losing sessions and rare large winning ones — the shape most likely to convince you it "almost works." The average of those shapes is unchanged.

How to read your results

Three tiles matter most. Bust risk is the fraction of 200 hidden runs where the progression demanded a bet the bankroll couldn't cover — for aggressive systems this number is the verdict all by itself. Worst losing run shows the streak your system must survive; compare it with how fast your bet sizes grow through a streak that long. And the final balance spread across shadow runs shows what the system really trades: every one of them has the same average, so the only thing you're choosing is the shape of the scatter around a −5.26% drift.

If a run finishes up, that's real — sessions end up all the time; that's why anyone gambles. Run the same system across 20 seeds and watch the average find its way home. The Lab isn't rigged against systems. Roulette is.

Why no staking system can beat the edge

One line of arithmetic: every bet you place has expected value of −(stake × 5.26%), regardless of what happened before, because the wheel has no memory. A staking system is a rule for choosing stake sizes — so a system's expected result is just the sum of negative numbers it chose. Bigger bets after losses, smaller after wins, elaborate lines of bookkeeping: all of it selects which negative numbers to add. There is no sequence of bad bets that sums to a good one. This is the same identity we walk through in our Stake Originals guides, where the house edge is at least printed on the screen.

What a system does control is the experience: how often you win a session, how bad the bad nights get, how long a bankroll lasts. Those are real choices worth making deliberately — as entertainment decisions, never as investments. If you take one number from this page: the cost of play is total amount staked × house edge, and turnover is the only lever you control.

Questions people actually ask

Does the Martingale system work on roulette?

No. Doubling after losses trades a high chance of small wins for a small chance of losing everything, with identical expected value to flat betting — every bet still carries the 5.26% edge. Table limits and finite bankrolls guarantee the eventual losing streak can't be covered; the Lab's bust-risk tile shows how often, and you can reproduce any run from its seed.

Is this simulator rigged to make systems lose?

It's seeded and reproducible — same seed, same spins, forever — and it uses the true double-zero probabilities (18/38 per even-money bet). Systems lose in it for the same reason they lose at casinos: arithmetic, not implementation. Export the CSV and audit any run spin by spin.

What's the best betting system for roulette?

None has a better expected value than any other — all of them lose stake × 5.26% per spin on average. The honest question is which session shape you prefer: flat betting for the longest, calmest ride; loss-chasing progressions for frequent small wins with rare disasters; win-pressing for frequent small losses with rare big nights.

Why double-zero roulette specifically?

It's the standard American wheel and a clean, well-known 5.26% edge — a fair, slightly tough test bench. A system that can't survive here can't survive anywhere, and one that could beat 5.26% would beat 2.7% single-zero too. No system does either.