Martingale: why it seems to work (until it ruins you)
Martingale is the most sold and least understood idea in this niche. It genuinely produces a streak of green days — which is exactly what makes it dangerous.
The math nobody shows you
At 87% payout, recovering a loss requires staking more than double — about 2.15x. Here is how the requirement grows on a losing streak starting at $10:
| Step | Required stake | Total exposed | Real-world frequency |
|---|---|---|---|
| 1st | $10 | $10 | Normal |
| 3rd | $46 | $78 | Happens every week |
| 5th | $214 | $391 | 3.1% of sequences get this far (0.42⁴, at 58% accuracy) |
| 7th | $988 | $1,838 | Happens a few times a year |
| 9th | $4,566 | $8,527 | Rare — and it's what wipes the account |
The ninth consecutive loss looks unlikely, but anyone trading 20 times a day meets long streaks regularly. That single event erases weeks of gains.
What our data shows
We tested martingale on the same validated strategies. Result: the long-run average does not improve — only the shape of the curve changes. With martingale the equity curve rises smoothly and drops in cliffs; without it, it wobbles more but never collapses. Across simulation grids, fixed stake plus a level-based daily target consistently beat aggressive martingale.
What we use instead
Fixed stake (1–2% of bankroll), a daily stop, and progression only between days, never within a day: level up only after a day that hits target; a losing day repeats the level. Across a 25-combination grid, target 3x with stop 5x was the most consistent point. It isn't magic — it caps the damage of bad days.
Most important: no money management rescues a losing strategy. Find the edge first (see the tests), then manage it.
Progression table: entry 1, factor 2.2, payout 85%
The bot's config_bot.ini ships with martingale = nao (off), martingale_fator = 2.2 and martingale_max = 2. At 85% payout, the exact factor that recovers previous losses plus the original profit is 1.85 ÷ 0.85 = 2.18, so 2.2 sits slightly above it: whichever level wins closes the cycle with a small profit. The ladder in units (multiply by your entry size):
| Level | Stake | Total staked so far | Net result if this level wins | Units at risk per unit of profit |
|---|---|---|---|---|
| Entry | 1.00 | 1.00 | +0.85 | 1.2 |
| Gale 1 | 2.20 | 3.20 | +0.87 | 3.7 |
| Gale 2 | 4.84 | 8.04 | +0.91 | 8.8 |
| Gale 3 | 10.65 | 18.69 | +1.01 | 18.5 |
| Gale 4 | 23.43 | 42.11 | +1.22 | 34.4 |
| Gale 5 | 51.54 | 93.65 | +1.69 | 55.3 |
| Gale 6 | 113.38 | 207.03 | +2.72 | 76.0 |
| Gale 7 | 249.44 | 456.47 | +4.99 | 91.5 |
| Gale 8 | 548.76 | 1,005.22 | +9.98 | 100.7 |
Read the last column: at gale 5 you have 93.65 units on the table to win 1.69. Profit barely moves while the amount at risk grows 2.2 times per level.
Capital needed for N gales — and how often the ladder breaks
Capital to survive N gales = entry × (2.2^(N+1) − 1) ÷ 1.2. The chance that one cycle loses every level is (1 − p)^(N+1). We show p = 57% (the middle of the out-of-sample range of our best strategies) and p = 50% (OTC, or any strategy without edge):
| Gales allowed | Capital, entry $1 | Capital, entry $10 | Whole ladder lost, 57% accuracy | Whole ladder lost, 50% |
|---|---|---|---|---|
| 0 (no martingale) | $1.00 | $10.00 | 43.0% | 50.0% |
| 1 | $3.20 | $32.00 | 18.49% | 25.0% |
| 2 (bot default) | $8.04 | $80.40 | 7.95% | 12.5% |
| 3 | $18.69 | $186.88 | 3.42% | 6.25% |
| 4 | $42.11 | $421.14 | 1.47% | 3.13% |
| 5 | $93.65 | $936.50 | 0.63% | 1.56% |
| 6 | $207.03 | $2,070.30 | 0.27% | 0.78% |
| 7 | $456.47 | $4,564.66 | 0.12% | 0.39% |
| 8 | $1,005.22 | $10,052.24 | 0.05% | 0.20% |
How often long losing streaks actually show up
The per-cycle chance looks small, but you run many cycles. Below is the probability of at least one run of K consecutive losses somewhere in 100 trades (about 5 days at 20 signals a day) and in 440 trades (20 signals × 22 trading days, roughly a month), computed exactly with a Markov chain:
| K losses in a row | 100 trades, 57% | 100 trades, 50% | 440 trades, 57% | 440 trades, 50% |
|---|---|---|---|---|
| 5 | 57.2% | 81.0% | 97.9% | 99.9% |
| 6 | 29.8% | 54.6% | 80.0% | 97.2% |
| 7 | 13.8% | 31.8% | 49.4% | 82.6% |
| 8 | 6.1% | 17.0% | 25.2% | 57.8% |
| 9 | 2.6% | 8.8% | 11.7% | 34.7% |
| 10 | 1.1% | 4.4% | 5.2% | 19.1% |
Six losses in a row is exactly what wipes a 5-gale ladder. Even with a genuinely good 57% strategy, it happens at least once in a 440-trade month with 80.0% probability.
Risk of ruin: one simulated month
Setup: bankroll 100 units, entry 1, factor 2.2, payout 85%, 440 signals, 200,000 simulated months per scenario (Monte Carlo). If the bankroll cannot cover the next stake, the account stops. "Flat 2.23" bets, on every signal, the same average amount that the 5-gale martingale ends up betting at 57% — a like-for-like comparison.
| Accuracy | Staking | Average final bankroll | Ends below 100 | Ends at 50 or less | Could not afford next stake |
|---|---|---|---|---|---|
| 57% | Flat 1 | 123.9 | 10.0% | 0.01% | <0.01% |
| 57% | Martingale, 2 gales | 142.1 | 14.9% | 1.38% | 0.17% |
| 57% | Martingale, 5 gales | 153.5 | 39.1% | 38.2% | 37.6% |
| 57% | Flat 2.23 | 153.5 | 10.1% | 0.79% | 0.08% |
| 50% | Flat 1 | 67.0 | 95.3% | 18.3% | 0.05% |
| 50% | Martingale, 2 gales | 42.6 | 91.6% | 60.3% | 29.6% |
| 50% | Martingale, 5 gales | 46.4 | 83.4% | 81.9% | 82.3% |
Two lessons. At 57%, martingale raises the average only because it puts more money on each signal: flat 2.23 units reaches the same average (153.5) with a 0.79% chance of ending at half the bankroll, against 38.2% with 5 gales. At 50%, martingale makes both the average and the bad tail worse.
Why martingale cannot create an edge
Each trade is a separate bet with the same odds. Every unit staked, at any level of the ladder, has the same expectation: e = p × 0.85 − (1 − p), which is +0.0545 per unit at 57% and −0.075 per unit at 50%. The ladder only changes how many units you stake after a loss.
Check it with the default 2-gale cycle (1, 2.2, 4.84). At 50%: expected result per cycle −0.248 units for 3.31 units staked on average, that is −7.5% per unit — identical to flat betting. At 57%: +0.155 for 2.841 staked, +5.45% per unit — again identical. Five losses in a row do not make the sixth trade more likely to win; its chance is still p.
The alternative in numbers: level-based daily target
The bot's plan (panel, "Risk" tab): a base stake (e.g. $50), a step (e.g. +$10), and the day ends at +3 stakes (target) or −5 stakes (stop). You level up only after a day that hits target; a losing day repeats the level; a level cap (e.g. 10) limits the stake to $140. The worst day is known in advance: about −5 stakes (the last losing trade can push it slightly past −5, never to −6), so roughly −$250 at level 1 and −$700 at level 10.
Computed with an absorbing Markov chain for one day at flat stake and 85% payout, running until target or stop:
| Accuracy | Day reaches +3 stakes before −5 | Average trades until the day ends |
|---|---|---|
| 50% | 43.2% | 21.5 |
| 54% | 61.4% | 21.4 |
| 57% | 74.0% | 20.0 |
| 60% | 84.1% | 17.9 |
Look at the 54% row: practically break-even (54.05%), yet 61.4% of days hit target. A high share of green days is not proof of profit, because a losing day costs 5 stakes and a winning day pays 3. The plan limits damage; the edge must come from a strategy measured out-of-sample (our tests). Try any setting in demo first — see the demo account guide.
🤖 Get the free trading bot
Open-source bot for Quotex & IQ Option (13 strategies). Edit it with ChatGPT/Claude — no coding. Free, no spam.
FAQ
Does martingale work in binary options?
It creates no edge. It raises your share of winning days and concentrates losses into rare, large events. With payout under 100%, expectancy stays negative if the strategy is negative.
How many martingale steps are safe?
No number is safe: each extra step multiplies exposure by ~2.15 (at 87% payout). What actually caps risk is a fixed stake with a daily stop.
Why do so many people recommend it?
Because it works beautifully for a few weeks — long enough to sell a course or film a testimonial. The account breaks later.
With the bot's default of 2 gales, how often does the whole ladder lose?
At 57% accuracy, in 7.95% of cycles (0.43³); at 50%, in 12.5%. Each full loss costs 8.04 entries, while a winning cycle nets 0.85 to 0.91 entries — so it takes about 9 winning cycles to pay for one full loss.
⚠️ Binary options carry a high risk of losing all your capital. Educational content — not investment advice. No money-management scheme turns a losing strategy into a winning one.