Martingale and grid trading in Expert Advisors: what they are, why they eventually fail and how to spot them

Martingale doubles the lot size after every loss so that a single winner wins it all back. A grid opens orders every few dollars, often without a stop loss, and waits for the price to come back. Both produce curves that climb in an almost straight line, and both hide the same thing: a rare but huge loss that can wipe out the account. Here’s the math on gold, and how to spot them before you pay.

What martingale is in trading, and its variants

The martingale strategy is a rule for sizing each trade: after a loss, the next trade opens with double the lot size. It comes from gambling: when you finally win, that one winner recovers everything you lost, plus the profit the first trade was aiming for.

An example on gold. With the usual contract of 100 ounces per lot (check it in your broker’s Specification window, as we explain in how to calculate lot size in gold), 0.01 lots make or lose $1 for every dollar the price moves. With the stop loss and the target both at $10, the first trade, at 0.01 lots, risks $10. If it loses, the second goes on with 0.02; if that loses too, with 0.04, and so on. After three losses, for example, $70 is already gone (10 + 20 + 40) and the fourth trade goes on with 0.08 lots:

Whenever the win comes, the cycle leaves $10 before costs. But after seven losses in a row, the eighth trade goes on with 1.28 lots, 128 times the first, and $1,270 is already gone. And margin stops you sooner: with gold at $4,000 and the 1:20 maximum leverage for retail clients in the EU and the UK, 1.28 lots tie up $25,600 (how it’s calculated, in margin call and stop out). To open that eighth trade you’d need almost $27,000 in the account. All of that, to make $10.

Soft martingale and recovery modes

A soft martingale multiplies by less, by 1.5 for example: after seven losses, the lot is about 17 times the starting size instead of 128, but from two losses in a row onwards, a single winner no longer wins it all back. Recovery modes open bigger or more numerous trades after a loss. The name changes, the idea doesn’t: the size grows with the loss.

What grid trading and averaging down are

A grid places orders at fixed price intervals: a buy every $10 that gold falls, a sell every $10 that it rises, or both. Many grids don’t use a stop loss: they close the whole group once the price has come back far enough for the total to be in profit.

Averaging down is the best-known case: you buy, the price falls and you buy again lower, so your average entry price drops and a smaller bounce is enough to get out. It’s a one-way grid, and if the lot size is multiplied at each level, it’s a grid with martingale. It almost never closes a trade at a loss; the problem is what happens if the price doesn’t come back.

Why the curve looks so good, and why it ends badly

Neither of them turns a losing strategy into a winning one: they change how the results are spread out. Many small losses become many small wins and one rare, huge loss. That’s why the balance curve climbs in an almost straight line, until the day it doesn’t.

Martingale: the streak comes sooner or later

Back to the example, now with a cap: the robot goes back to 0.01 lots after a win or after eight losses in a row. With a 50% win rate and no costs, a cycle loses all eight once in every 256 (0.5 to the power of 8). The other 255 make $10 each, $2,550 in total; that one loses $2,550 in one go. The expectancy, what each cycle makes on average, is exactly zero, and with spread and commissions it’s negative: martingale doesn’t create an edge, it concentrates the risk in one streak.

When does it come? Nobody knows. By our calculation, the probability of having hit it reaches 50% around cycle 177, after about $1,770 of very steady gains, and 98% over 1,000 cycles. And over 1,000 trades with a 50% win rate, you’ll most likely see at least one run of eight losses in a row (an 86% probability): streaks run longer than we expect, as we explain in losing streaks in trading.

Grid: gold moves more than it seems

In a grid, the loss stays open and grows faster than the move: every dollar the price falls is lost by all the positions at once. Here are two buy grids with no stop loss against a fall in gold; each cell shows the open lots and what they’re losing together (with the fixed grid, after a $100 fall, that’s 11 buys of 0.01: the first one and one every $10):

In the fixed grid, if the fall goes from $100 to $200, the loss almost quadruples. The one that doubles the lot is already holding 2.55 lots after a $150 fall: about $51,000 of margin at 1:20 with gold at $4,000. In return, at that point a bounce of about $29 is enough for it to close without a loss, against $75 for the fixed grid: that’s why it almost always works out.

And falls like these aren’t rare. According to Dukascopy’s public data, between August and October 2026 gold moved more than $100 from high to low on 21 of its 65 trading days; on Friday, August 28 it fell $185.63 in under five hours, and from the August 25 high ($4,696.64) to the October 30 low ($3,869.98) it dropped almost $827 (more data in the best timeframe to trade gold). A grid without a stop has no set loss limit: the money in the account sets it, and what cuts it off is the stop out, when the broker force-closes positions.

That also explains the win rate: in a grid it shoots up, because losing trades aren’t closed, they wait; in a martingale with a stop, what looks close to perfect is the share of cycles that end in profit. It’s the pattern in why a high win rate can be a trap.

How to spot martingale or grid in a MetaTrader 5 backtest

A Strategy Tester backtest leaves traces even when the balance curve is perfect (how to run one, in how to backtest in MetaTrader 5):

• Equity drifts away from balance. On the Graph tab, the blue line is the balance and the green one is equity (the balance plus or minus what open trades are making or losing). If the green line drops well below the blue one and then comes back, there are open losses the balance isn’t showing.
• A much deeper drop in equity. In the report on the Backtest tab, “Equity Drawdown Maximal” is usually somewhat larger than “Balance Drawdown Maximal”, because it counts open trades; if it’s several times larger, the robot is holding on to losing positions (more in what is drawdown).
• A high deposit load. The Deposit Load histogram underneath measures margin against equity: tall spikes mean many positions or big lots at the same time.
• A very high win rate, huge average losses. A very high “Profit Trades (% of total)” and an “Average loss trade” many times bigger than the “Average profit trade”. “Maximum consecutive losses ($)” shows the longest losing streak and, in brackets, what it cost; “Maximal consecutive loss (count)” shows the streak that lost the most money.
• Lot sizes that grow or pile up. “Open Chart”, in the context menu of the Backtest tab, plots all its trades on the price chart, and from the same menu, usually under “Report”, you can save the results as HTML or Excel (Open XML), with the list of deals and their volume. Buys stacked lower and lower, or lots that double after a loss, are the signature.
• Trades without a stop loss. If positions don’t carry one from the moment they open, everything depends on the price coming back (we cover it in stop loss, take profit and break-even).

A grid can come out of a backtest without big drawdowns if the period didn’t include the move that breaks it: how it handles a loss matters more than the curve it drew (backtesting and overfitting).

How to spot martingale or grid on Myfxbook

On an account published on Myfxbook, look in these places (what each figure means, in how to read a Myfxbook account):

• Lots, in the History tab. Lot sizes that double after a loss, or several positions in the same direction opened at stepped prices and closed together.
• Equity versus Balance. In Stats, equity well below balance means open losses; on the Growth chart, the Equity Growth line pulling away from Growth shows it over time. Myfxbook’s Drawdown already includes floating losses.
• Worst Trade versus Average Win. In Advanced Statistics, a worst trade worth dozens of average wins, with a very high win rate, is the fingerprint of a grid or a martingale.
• The Streak tab. Longest Losing Streak is the longest run of losses, in trades, and Worst Losing Streak, what the worst run cost. In a martingale, a few losses in a row cost many average wins.
• Open Trades. If it’s public, it shows the open positions. Many accounts hide it, which is common practice so trades can’t be copied in real time; equity still tells you how they’re doing.

Clues in an EA’s inputs, and what to ask the seller

An Expert Advisor’s settings are on the Inputs tab. Each developer chooses the names, so they’re only a guide: what gives the technique away isn’t the name, but whether the lot size grows after a loss or orders open at a fixed distance from the previous one. These names often go with it:

• LotExponent or Martingale: how much the lot size is multiplied by after a loss, or with each new order in the series. A multiplier applied to a fixed lot you choose yourself is something else: that lot doesn’t change after a loss.
• PipStep or GridStep: how far from the previous order the next one in the series opens.
• MaxOrders or MaxTrades: how many orders it can have open at once. If it allows several in the same direction, it can stack them; a cap on trades per day is something else.
• Grid, Averaging or Recovery: they usually switch on a grid, averaging down or a recovery mode; check whether they’re enabled.

And before you pay, ask in writing:

• What does the robot do after a loss: keep the lot size, reduce it or increase it?
• Can it have several positions open at once on the same symbol? How many?
• Does every trade carry a stop loss from the moment it opens?
• What has been the largest open loss, not just the drop in balance?
• How much margin does it need in the worst case, with my capital and my leverage?

Legal but risky: martingale, grid and prop firms

Martingale and grid aren’t illegal in themselves: they’re ways of deciding the size and number of trades, and some traders use them knowing exactly what they’re doing. What they lack is a small maximum loss known in advance: either there’s no cap (a grid without a stop, a martingale with no limit on the number of steps) or, if there is one, it’s huge compared with what each cycle makes. What’s misleading is selling them without saying so (more in red flags of a trading robot scam).

Several prop firms ban them explicitly in their rules, so read your firm’s rules before you run an EA. Others don’t name them, but their daily and maximum loss limits clash with any technique that piles up open losses (we cover it in prop firm rules for an EA).

Checklist: does this robot use martingale or grid?

• Lot sizes don’t grow after a loss, either in the backtest or on the account.
• There are no stacked positions in the same direction at stepped prices.
• Every trade carries a stop loss from the moment it opens.
• The maximum drawdown on equity isn’t several times the one on balance.
• The account’s equity doesn’t sit well below its balance.
• If it wins almost every time, its worst trade isn’t worth dozens of average wins.
• No input increases the lot size after a loss or opens orders in a series; if one looks like it from its name, the seller explains in writing what it does.
• The seller answers in writing what the robot does after a loss.

Passing the list doesn’t make a robot profitable: it only tells you its risk is out in the open. The rest is in what to check when choosing an Expert Advisor for gold.

AUREUM: no martingale, no grid

AUREUM is an Expert Advisor for gold (XAU/USD) on MetaTrader 5, and it doesn’t use martingale or grid: it doesn’t increase the lot size after a loss or open trades to average down against the move. Every trade opens with its loss and profit limits already calculated. With its reference lot mode, which is how it comes set up, the size is 0.01 lots per 1,000 of your account balance, rounded to the step your broker allows, with a minimum (your broker’s minimum lot) and a cap of 1 lot. Because it depends on the balance, after a losing streak the lot size stays the same or goes down; it never grows to win back what was lost. If you see mmMultiplier in its Inputs, it only counts in manual mode (LOT_MANUAL), where the lot is mmLots × mmMultiplier: a fixed size you choose that doesn’t change after a loss.

Its published backtest, a simulation on XAU/USD on the 1-hour chart (H1) from 2020 to 2026, has 439 trades and a maximum drawdown on balance of 21.84%. The home page also has a live account connected to Myfxbook.

Frequently asked questions about martingale and grid