Martingale vs Fixed Lot on the Same Entries: Doubling Broke on 57 of 66 Start Dates, and Only Paid Where the Fixed Lot Already Won

Martingale is the oldest money trick in trading: if you lose, double the next bet, and the first win recovers everything plus the original profit. Many free robots use it quietly. Their inputs say “Multiplier 2.0” or “LotExponent”, and their results show a long, smooth rising line.

The question people ask is whether martingale is better than a fixed lot. The fair way to answer it is to hold everything else still. So I used one entry rule and traded it both ways on the same bars. The short answer: martingale did not change whether the entry made money. It changed how fast, and in the losing cases, how the account ended.

How was the comparison set up?

• Data: 16,925 EURUSD M5 bars, June 3 to August 26, 2026, split at July 15 into two halves.
• Entry: in the direction of the last closed bar, at the open of the next one. One trade at a time. This is deliberately a plain rule with no edge of its own, so any difference comes from the lot size.
• Exits: a fixed target and stop. Four pairs: 5/5, 10/10, 20/20 and 10/20 pips. If a bar touched both, the stop counted.
• Lot rules: fixed 0.01; martingale ×2 (lot doubles after a loss, resets to 0.01 after a win); ×2 capped at 5 doublings, then back to 0.01; and a softer ×1.5.
• Account: $1,000, 1:100 leverage. If the next lot needs more margin than the account holds, the robot stops. That is what I call “broke” below.
• Cost: 1 pip per trade, from my spread and swap test.
• Start dates: one run from June 4, plus one from each of 66 trading days.

What happened with a 10-pip target and stop?

This is the classic martingale setup, and the entry lost money on its own:

Run from June 4 unless stated. “Broke” and “Profitable” count the 66 start dates.

A 42.2% win rate with equal target and stop is a losing rule: the pip of cost means each winner needs an 11-pip move and each loser only 9. The fixed lot lost slowly and evenly, $46.00 in the first half and $36.27 in the second. It lost on every one of the 66 start dates, never by much.

The pure ×2 martingale lasted 27 trades. On June 8 it lost 7 in a row, and the eighth trade required 1.28 lots. It stopped with $884.63, down more than the fixed lot managed to lose in twelve weeks. Across the 66 start dates it broke 57 times.

The capped version is the one many robots actually use, because it “limits the risk”. It did survive every start date. It also lost the most: every time a streak reached 6 losses, 5 doublings after the first trade, it booked the whole ladder, 0.01 + 0.02 + 0.04 + 0.08 + 0.16 + 0.32 lots, and started again.

How long are the losing streaks?

This is the number martingale depends on. With the fixed lot, 538 trades, 10/10:

Streak counts from the fixed-lot run. Margin at a EURUSD price of 1.15. A 10-pip loss costs $1 on 0.01 lots, $2 on 0.02 and so on, so a streak of n losses costs $2ⁿ − 1.

The dollar damage of the streak is not what stops a small martingale: seven losses cost $127. The margin is. A $1,000 account at 1:100 can hold the 0.64-lot step and cannot hold the 1.28-lot one. Seven-loss streaks came up 7 times in twelve weeks.

With higher leverage the ladder goes a step or two further, and the account has more to lose when it finally ends. The streaks themselves do not get shorter. My position sizing table shows what a fixed risk per trade looks like instead.

Did martingale ever beat the fixed lot?

Yes, and only when the fixed lot was already winning. Here are the other target and stop pairs:

End balance from June 4, $1,000 start. “(broke)” means the robot stopped when it could not afford the next lot.

The two top rows are losing entries, and martingale made them worse or ended them early. With 5-pip targets the ×2 robot broke after 17 trades, half a day, and on all 66 start dates. The ×1.5 broke on all 66 too. Their end balances look better than the fixed lot’s only because they stopped trading. The capped version kept going and finished at $372.50.

The two bottom rows are entries that made money with a fixed lot in this sample: 157 trades with 20/20 made $16.14 in the first half and $22.88 in the second. Here the martingale ×2 made $191.73, almost five times the fixed lot, and never broke. The reason is visible in the largest lot: 0.16. With 20-pip stops the longest losing streak was 4 trades, so the ladder never reached the rungs that need real margin.

So martingale is an amplifier, not an engine. It took a small positive result and made it larger, and it took a small negative result and turned it into a broken ladder. It could not tell the two apart, and neither can you from a win rate. The 20/20 rule won 54.8% and the 10/20 rule 67.1%, and the higher win rate made less money.

Is a small edge in twelve weeks enough to trust martingale?

No. The 20/20 result is 157 trades on one pair over one summer. My in-sample vs out-of-sample test showed how often a setting that works in one period stops working in the next, and my optimised vs default settings test found the same thing for indicator inputs. If the edge disappears, the martingale is back in the top rows of the table, with a larger balance to lose.

A fixed lot fails gracefully. When the 10/10 rule stopped working, the fixed-lot account lost 8.2% over twelve weeks and you would have had months to notice. The martingale version found out in 4.2 days.

So which should you use?

• Use a fixed lot, or a fixed risk per trade, to find out whether an entry works. Martingale hides the answer: the curve looks smooth until the day it doesn’t.
• Never use martingale on an entry that loses with a fixed lot. In every losing setup here it either lost more or stopped the robot within days.
• A cap on doublings does not make martingale safe. It turns a rare large loss into regular medium ones. The capped 10/10 version lost $290.18, three and a half times the fixed lot.
• If a free robot has a “Multiplier” or “LotExponent” above 1.0, it is a martingale or a grid. Check what it does in my test of five free robot logics and set it up with the numbers from the grid EA settings guide.
• Stops matter more than the lot rule. My stop-loss distance test and fixed vs ATR stop comparison are where to start.

Frequently asked questions

Is martingale better than a fixed lot?

Only when the entry already makes money. With a 20-pip target and stop, the fixed lot made $39.02 and the martingale $191.73. With a 10-pip target and stop, the fixed lot lost $82.27 and the martingale ran out of margin on 57 of 66 start dates.

How many losses in a row can a martingale survive?

On $1,000 with 0.01 lots at 1:100, seven. The trade after the seventh loss requires 1.28 lots, about $1,472 of margin, and the robot cannot open it. In 538 trades, streaks of seven or more losses happened 7 times.

Does capping martingale at 5 steps make it safe?

It stopped the account from breaking, and it lost more. With a 10-pip target and stop the capped version ended at $709.82, the fixed lot at $917.73.

Is a 1.5 multiplier safer than 2.0?

Slower, not safer. With a 5-pip target and stop the ×1.5 martingale broke on all 66 start dates, the same as ×2.

Why do martingale robots show such smooth equity curves?

Because almost every streak ends in a win that recovers it. The curve rises in small steps until a streak is longer than the account can pay for. In this test that took a median of 6 days with 10-pip stops.