No betting system beats the house edge in crash. Changing your stake after wins or losses changes the shape of your results (more small wins, rarer huge losses, or the reverse), but every unit you wager still carries the same expected loss. We simulated four common systems to show what that looks like in practice.
This guide explains each system, shows the simulation, and walks through the martingale ladder that catches most players out.
The four systems tested
All four use a 2.00x auto-cashout (about a 49.5% win chance at 1% edge) and a base stake of 1 unit:
- Flat: stake 1 unit every round, no matter what.
- Martingale: double the stake after each loss; return to 1 after a win. The idea is that one win recovers all previous losses plus 1 unit.
- d'Alembert: add 1 unit after a loss; subtract 1 after a win (minimum 1). A gentler version of "raise after losses".
- Paroli: double after a win, up to three wins in a row (1, 2, 4), then reset; reset after any loss. The reverse idea: raise after wins.
The simulation setup
We wrote a Python simulation using a standard crash formula, bust = 0.99 ÷ (1 − R) rounded down to two decimals with a floor of 1.00x, which gives a 1% house edge. Each session:
- starts with a bankroll of 100 units;
- plays up to 200 rounds;
- stops early if the bankroll can't cover the next stake the system requires.
We ran 20,000 sessions per system with a fixed random seed so the results can be reproduced. These are simulated sessions, not records of real play.
The results
| System | Average final balance | Median final balance | Sessions finishing up | Sessions finishing at or below 50 | Sessions stopped early (couldn't cover next bet) |
|---|---|---|---|---|---|
| Flat | 97.96 | 98.0 | 41.6% | 0.1% | 0.0% |
| Martingale | 94.94 | 55.0 | 33.9% | 43.8% | 66.0% |
| d'Alembert | 90.54 | 63.0 | 47.0% | 49.4% | 48.3% |
| Paroli | 96.49 | 96.0 | 43.4% | 4.5% | 0.0% |
Starting balance in every case: 100 units.
What the numbers say:
- Every system finished below 100 on average. None turned the −1% edge into a gain.
- Flat betting lost the least. It wagered 200 units per session, and 1% of 200 is 2: an average finish of about 98 is exactly what the math predicts.
- Martingale and d'Alembert lost more, not less. They raise stakes after losses, so they wager more in total. They also hit the bankroll wall: once a run of losses makes the next bet unaffordable, the loss is locked in.
- "Finishing up" is a misleading score. d'Alembert finished up in 47% of sessions, the highest of the four, and also had the worst average. Lots of small wins, paid for by fewer, much larger losses.
The martingale ladder
Martingale fails in a predictable way. At a 2.00x target each round loses about 50.5% of the time. Here is what a losing streak costs, starting at 1 unit:
| Losses in a row | Next stake | Total lost so far | Chance of this streak from any start |
|---|---|---|---|
| 1 | 2 | 1 | 50.500% |
| 2 | 4 | 3 | 25.503% |
| 3 | 8 | 7 | 12.879% |
| 4 | 16 | 15 | 6.504% |
| 5 | 32 | 31 | 3.284% |
| 6 | 64 | 63 | 1.659% |
| 7 | 128 | 127 | 0.838% |
| 8 | 256 | 255 | 0.423% |
| 9 | 512 | 511 | 0.214% |
| 10 | 1,024 | 1,023 | 0.108% |
With a 100-unit bankroll, six straight losses use up 63 units and the seventh bet would need 64. You can't place it. A 0.838% chance per starting point sounds small, but you start a new streak after every win. In our simulation, about 57% of 200-round sequences contained at least one run of seven or more losses.
So the typical martingale experience is: dozens of rounds of +1 unit, then a single streak that costs 63 units or more. The average of those two experiences is negative, by exactly as much as the extra amount wagered times the edge, plus the cost of being forced to stop.
Worked example: one fictional martingale session
The session below is invented for illustration (bankroll 100, base 1, target 2x):
| Phase (fictional) | Rounds | Result | Balance after |
|---|---|---|---|
| Steady wins and short streaks | 1–60 | +31 net (31 completed cycles) | 131 |
| Six-loss streak begins | 61–66 | −1, −2, −4, −8, −16, −32 | 68 |
| Seventh stake would be 64 | 67 | Win at 2x: +64 | 132 |
| Another streak, this one seven long | 68–74 | −1 … −64 (total −127) | 5 |
| Next stake would be 128 | 75 | Can't cover. Session over | 5 |
Nothing unusual happened. Two long losing streaks in 75 rounds is well within normal variation. The first one happened to break just in time; the second didn't.
Why systems can't work
The reason is simple enough to state in one line: each round's expected value is −edge × stake, and adding up rounds never makes a sum of negatives positive. A system can only choose how big each stake is. Bigger stakes after losses mean more total wagering and a bigger expected loss. Bigger stakes after wins (Paroli) also mean more wagering. Nothing about the order of results changes the expected value of the next round.
Checklist: spotting a system pitch
When you see a "crash strategy", look for these:
- Claims it "can't lose" or "always recovers". Martingale recovers only if your bankroll and the site's maximum bet allow every doubling.
- Results from one session or a screenshot. One session shows variance, not value.
- "Wait for a streak, then bet". Past rounds don't affect future ones.
- No mention of total amount wagered. That number drives expected loss.
- Profit promises. No honest source can promise profit from a negative-expectation game.
If you play crash, flat stakes with a fixed limit make the cost easiest to see and control. For why no target helps either, see auto-cashout targets and expected value. If losses are something you find yourself chasing, our responsible gambling page lists free support.
FAQ
Does martingale work on crash games?
No. Martingale turns many small wins into rare, very large losses. In our simulation, martingale sessions finished with a lower average balance than flat betting, and 66% of them ended early because the bankroll could not cover the next doubled bet.
Why do martingale players often report winning?
Because most short sessions do end slightly up. The losses come from occasional long losing runs that wipe out many sessions' worth of small wins. Over enough sessions the average is negative.
Is there any betting system that beats the edge?
No. Each round has a negative expected value per unit staked. Changing stake sizes changes how much you wager and when, but cannot make the total expected value positive.
Why did some systems lose more than flat betting?
Systems that raise stakes after losses wager more in total, and expected loss grows with total amount wagered. They also hit bankroll limits, which locks in losses.