How much capital does a grid EA need? Sizing from the worst basket, not the average month

The question I get most often about a grid Expert Advisor is "what is the minimum deposit?". The honest answer is that the minimum deposit and the capital the strategy needs are two different numbers. The first one is whatever lets the terminal open a 0.01 lot. The second one is what keeps the account alive through the worst basket. This post shows how I work out the second number, with plain arithmetic and with the backtest figures of my two grid EAs, Snapback Pro and Fairway.

Why a grid needs more money than it seems to

A grid adds positions while price moves against it. Each new order is small, so the first days look harmless. The problem is that the floating loss does not grow in a straight line. Every further step loses money on all orders that are already open.

Take the simplest case: a fixed 0.01 lot, one order every 20 pips, and a pair where one pip on 0.01 lot is worth about 0.10 USD. Measured at the moment each new order opens:

Twice the orders means roughly four times the loss. This table is arithmetic, not a backtest, and it is the friendly version: the lot never grows. As soon as the lot increases with the depth of the grid, the curve gets steeper. A multiplier (martingale) makes it exponential, which is why I do not use one in any set file.

The second thing the table does not show is margin. Thirty open orders of 0.01 lot are a 0.30 lot position. With low leverage, the margin for that position is locked while the floating loss eats into equity, and the broker's stop-out level is reached earlier than the plain loss figure suggests. Look up margin and stop-out level in the contract specification of your account before you size anything.

The number that matters: the worst basket

For capital planning I ignore the average month and the profit factor. I look at one figure: the largest equity drawdown of a long test with real ticks, in money, at the lot size I plan to trade.

A trick that helps here is the measuring run: test with a fixed start lot and a very large balance. With a fixed lot the trades do not depend on the balance, so the large balance only makes sure that no stop out cuts the test short. You then see the full drawdown in money. A test on a small balance that ends in a stop out tells you only that the balance was too small, not how much would have been enough.

Example 1: Snapback Pro, a grid without a hard stop

Snapback Pro is a multi-pair reversal grid. Single positions have no stop loss and the supplied set files use no equity stop. Protection comes from capital, from the limit of 5 pairs at risk at the same time and from the basket trailing. So capital is the main safety setting.

These are the backtest figures from the guide (MT5 Strategy Tester, every tick based on real ticks, RoboForex ECN data, 1 Jan 2025 to 9 Oct 2026, fixed 0.01 start lot). They are backtests, not live results:

The rule behind the recommendation: the largest drawdown of the test should stay at or below roughly 35 % of the account. Two details are worth a look.

1. More pairs do not automatically mean more drawdown. The 5-pair set had a deeper drawdown than the 7-pair set, because only 5 pairs can be at risk at once and the result depends on which pairs hold those slots.
2. Too little capital ends the test. All 20 pairs on 3,000 USD reached a stop out in March 2026 in the test. The same logic with enough capital survived that period.

My own live Snapback account started with 1,000 USD, which is below the 2,500 USD I recommend for the 3-pair set. I say so in the signal description. Please do not read it as a recommendation to do the same.

Example 2: Fairway, a grid with a basket stop

Fairway trades AUDCAD on M5 and closes a basket completely when it reaches a set loss. In the default set that is 150 USD per 3,500 USD of balance, about 4.3 % per stopped basket. Lot and basket stop scale together in full steps of 3,500 USD.

Here the capital question changes. The loss of one bad trend is known in advance, so you do not have to hold capital for an open-ended basket. What you have to plan for instead is several stops in a row. Backtest from the guide (real ticks, RoboForex, AUDCAD M5, 3,300 USD, 1 Jan 2024 to 17 Oct 2026, default set):

The stop has a price. It turns some baskets that would have recovered into real losses, and 2024 ended slightly negative in this test because three baskets were stopped. A profit factor of 1.16 is a thin edge. But the drawdown figure you size for is 9 % instead of 57.8 %.

One point for small accounts: below 3,500 USD Fairway still trades 0.01 lot, so the 150 USD basket stop becomes a larger share of the account. On 1,500 USD it would be 10 % per stopped basket. Reduce the basket stop to match, or accept that the risk per basket is higher.

A simple way to work out your own number

1. Run a test with every tick based on real ticks on your broker's data, over at least two years, with the set you plan to use.
2. If the EA has no hard stop, repeat it as a measuring run: fixed lot, very large balance. Note the largest equity drawdown in money.
3. Decide which share of the account you can watch being under water without switching the EA off at the worst moment. For most people that is less than they think.
4. Divide the drawdown by that share. 818 USD at a 35 % tolerance is about 2,340 USD, which is where the 2,500 USD of the 3-pair set comes from. At a 20 % tolerance the same set needs about 4,100 USD.
5. Add a margin for the future. A backtest covers the past, and the largest drawdown of a live account usually comes later. How much you add is your decision; my own habit is to treat the tested drawdown as the lower bound, not as the limit.
6. Check margin and stop-out level for the deepest basket of the test on your leverage.
7. Start on a demo account, then live with the smallest lot.

What capital does not fix

More capital makes a stop out less likely. It does not make the strategy better. A grid without a stop that meets a trend longer than anything in the test period can still take a large part of the account, only later. If you want a hard limit, use an equity stop or a basket stop and accept that it will turn some drawdowns into realised losses. Swap also counts: baskets can stay open for days, including the day with triple swap.

Trading forex and CFDs on margin carries a high risk of loss. Backtests are simulations and past results do not guarantee future results. Please start on a demo account.

Snapback Pro guide with all set files: https://www.mql5.com/en/blogs/post/778006

Fairway guide: https://www.mql5.com/en/blogs/post/778374

All my products: https://www.mql5.com/en/blogs/post/778024