20 systems · 20,000 identical seeds each

Does any roulette betting system work?

20 betting systems, each run through the same 20,000 seeded sessions on a European wheel and again on an American one. The most persistent walks away $200 up 78% of the time. Every one of them loses between $2.52 and $2.87 per $100 wagered, which is the house edge and nothing else.

Same rules, same spins, every system

Each session starts with $1,000, bets in $5 units, and ends at the first of: up $200, 500 spins, or a required stake the bankroll or the $500 per bet limit can’t cover. Every system sees the identical sequence of spins, so differences between rows come from the betting rules alone.

Betting systems ranked by how often they walked away $200 up, European wheel
SystemWalked away +$200BustedMean resultPer $100 wagered
Dozen Martingalenegative progression77.8%22.2%−$29−$2.55
Labouchèrenegative progression76.2%23.8%−$50−$2.69
D’Alembertnegative progression71.6%28.4%−$119−$2.87
Oscar’s Grindnegative progression71.6%28.3%−$123−$2.67
Reverse Labouchèrepositive progression71.5%28.5%−$92−$2.66
James Bondcoverage69.7%30.3%−$136−$2.66
Fibonaccinegative progression69.1%30.9%−$72−$2.78
Martingalenegative progression68.7%31.3%−$32−$2.52
Grand Martingalenegative progression67.9%32.1%−$25−$2.66
Contra D’Alembertpositive progression67.8%31.4%−$160−$2.71
Single numberflat stake67.0%11.0%−$32−$2.74
Hot number (Andrucci-style)pattern66.2%9.6%−$30−$2.73
Reverse Martingalepositive progression65.2%23.0%−$96−$2.70
Fixed fraction (2%)flat stake41.8%0.0%−$142−$2.69
1-3-2-6positive progression29.2%0.0%−$111−$2.71
Parolipositive progression21.0%0.0%−$101−$2.69
Romanoskycoverage11.8%14.3%−$483−$2.70
Two Dozenscoverage5.5%0.0%−$133−$2.72
Flat bettingflat stake2.1%0.0%−$67−$2.71
Betting against a streakpattern0.0%0.0%−$4−$2.77

European wheel, 20,000 sessions per system. American-wheel results are on each system’s page. Download every row for both wheels (CSV).

Why the last column barely moves

Every bet on a European layout returns −1/37 of its stake on average, about −$2.70 per $100. A system decides when and how much to bet, but each of those bets still carries −1/37, so the total is −1/37 of everything wagered. The spread in the last column is sampling noise. What a system can change is the shape: how often a session ends ahead, how deep the losing sessions go, and how much gets wagered along the way. The same-seed comparison study works through this for five of them in more detail.

Negative progression

Raise the stake after losses. High success rates, paid for by rare sessions where the stake outgrows the bankroll or the table.

Positive progression

Raise the stake after wins. Small, frequent losses between streaks; the risk per spin stays low.

Flat stake

No progression. The control group: what the edge does with nothing on top.

Coverage

Spread one stake across many pockets. Win most spins, win small, lose big on the rest.

Pattern

Bet on what the wheel has been doing. On independent spins, the past says nothing about the next result.

Method

Sessions
20,000 per system per wheel, 800,000 in total
Master seed
0x6d2e6c8b, shared by every system
Outcome path
The two-stage seeded generator used by Engineer mode, unmodified
Verification
Each system’s Lua is replayed through the production kernel in the test suite and must match the bulk run exactly

Have a system that isn’t here? Write it in Engineer mode, run it on the same seeds, and send it in. If it beats the last column, that’s a bug worth finding.