Is Roulette Rigged Against Me? Checking Every Spin Since Launch
The draw picks one of 37 pockets at random and never reads who you are or what you bet. The real odds of each bet, a check of all 147 spins since launch, and what you cannot yet verify yourself.
Learn by playing: open Roulette in your browser, no download needed.
Play RouletteNo. The draw picks one of the 37 pockets, 0–36, with equal probability. It does not read who you are, what you bet on, how much you staked, your balance or your recent results. If it feels like nothing ever hits, that is usually just roulette: a straight-up bet on one number wins 1 time in 37, about 2.7% of spins. Here is how the number is chosen, followed by a check against every real spin since launch.
How the number is chosen
When you tap “Spin wheel”, the server checks your bets and balance, then draws a number from the operating system’s cryptographic random number generator. That line takes no input at all: no account, membership, bet area or stake goes into it, and there is no “help new players”, “losing-streak pity” or “pay out less on big bets” branch. Once the number is drawn, the return is computed from a fixed multiplier table and written as one ledger entry. The spinning wheel on screen is animation; the number is fixed before it starts.
Retrying or refreshing the same bet returns the saved result of that spin and never draws again. So the server cannot “redraw until you lose”, and you cannot “refresh until you win”.
Why it feels like nothing hits
| Bet | Pockets covered | Win chance per spin | Return on a win (incl. stake) | Long-run average return |
|---|---|---|---|---|
| Straight-up number | 1 | 2.7% | ×36 | 97.3% |
| Dozen or column | 12 | 32.4% | ×3 | 97.3% |
| Red/black, odd/even, low/high | 18 | 48.6% | ×2 | 97.3% |
Three things are easy to misread:
- A single number is meant to be hard to hit. Back one number for 37 spins and there is a 36% chance it never comes up; after 100 spins there is still a 6.5% chance.
- Even-money bets lose in streaks too. Because of the 0, red loses 51.4% of the time. Over 100 red bets, the chance of at least one run of 5 losses is about 85%, and of 7 losses about 36%.
- No betting pattern beats the zero. Every row above returns 36/37, or 97.3%, on average. Covering more numbers makes you win more often and less each time; the average stays the same. When 0 comes up, every outside bet loses and only a straight bet on 0 pays. The wheel has no memory, so the last spin says nothing about the next one.
Checking the real spins
We took every Roulette spin from launch on 3 November to the morning of 5 November (Singapore time): 147 spins from 26 players. Account details were dropped on export; only the bets, stakes, winning number and return were kept.
- Every settlement checks out: recomputing each return from the recorded bets and number matches what was credited on all 147 spins.
- The numbers look normal: each number came up about 4 times on average. The chi-square value is 31.5 (36 degrees of freedom; the normal range is roughly 21–54). Number 7 never came up, but over 147 spins it is common for one number to miss entirely.
- Slightly fewer paying spins, within normal variation: given the pockets each spin actually covered, 63.5 spins were expected to pay something back (at least one bet hitting) and 57 did.
- The closest thing to “big bets never win”: the 72 spins staked above the median (800 beans) were expected to pay back 32.4 times and did 24 times; one player’s 12 spins were expected to pay back 5.8 times and did twice. These two groups overlap heavily: that player’s stakes make up more than 70% of the high-stake group. They are the most striking gaps we found, but we picked them after looking at the data. We looked at 10 groupings in total: all spins, the smaller and larger halves of stakes, bets of 10,000 beans or more, and the 6 most active players. Simulating the same bets 20,000 times on a fair wheel, at least one grouping fell as far below expectation as the worst one here 13% of the time. Splitting the spins into earlier and later halves, or into two alternating groups, also stayed within normal variation.
- Total return was only 53.7%, mostly because of two spins: 171,000 beans on red that landed on black 10, and 157,750 beans on black that landed on red 23. Each was a 48.6% chance. Without those two, the other 145 spins returned 95.8%. Simulating the same bets on a fair wheel, 14% of runs return 53.7% or less. A handful of large stakes is enough to make short-term results swing hard.
Bottom line: we found no sign that the draw targets any player or bet type. Still, 147 spins is a small sample. The data can show that nothing looked wrong, but it cannot prove fairness by itself; the fairness comes from a draw that never reads your bets.
What you can verify
Roulette draws the number after the server has received your bets. Unlike TOTO, which commits to its draw in advance, Roulette has no draw proof yet, and its draw code is not in the public source repository. So “the draw never reads your bets” rests on our description of the code we run plus this statistical check; you cannot yet verify an individual spin yourself. What you can confirm: every spin’s bets, number and return are saved, refreshing or retrying never charges twice, and the multipliers and zero rule match the Roulette page. Beans are in-game points with no cash value; they cannot be transferred or withdrawn. If you want to lose more slowly, lower your stake per spin and use all-in less.