Grid Trading Explained: How Grid EAs Work, What the Maths Says and Where They Break

Two things called grid trading

The word covers two designs that behave very differently, and you should know which one an EA uses before you read anything else about it.

When people warn about grids, they mean the second kind. The first is closer to a breakout strategy with pending orders and carries the risks of one; our scalping guide covers its execution issues. The rest of this guide is about the averaging grid, because that is where the arithmetic gets interesting.

How an averaging grid works

The EA opens a first buy. If price falls by a set distance, say 20 pips, it opens a second buy at the lower price. If price falls another 20 pips, a third, and so on. Each new position pulls the average entry price down. As soon as price rises back to a point just above the average entry, the EA closes every position together at a small net profit. Because price in a range keeps returning to its average, most baskets close at a profit, quickly. That is where the 70 to 90 percent win rate comes from, and why a grid backtests so well on a calm market.

The problem is the right-hand panel. A grid has no opinion about whether the market is ranging or trending. It keeps averaging down into a trend, because that is what its rule says, until it runs out of margin, hits a stop or stops adding. All the profit it earned from the left-hand baskets is at risk in the right-hand one.

The arithmetic of a deep grid

Take EURUSD, a grid with a 20-pip spacing and a fixed lot of 0.01, where one pip on 0.01 lot is worth about $0.10. Suppose price falls steadily. When the grid has n levels open and price is at the bottom of the last one, the first position is n minus 1 spacings in loss, the second n minus 2, and so on. The total floating loss is the spacing times the pip value times n(n-1)/2, so it grows with the square of the number of levels.

Look at what quadrupling the depth from 10 to 40 levels does: the loss goes from $90 to $1,560, about seventeen times, while margin goes up four times. On a $1,000 account the grid at 40 levels is long past a margin call at 50 percent margin level, which is the close-out level for European retail accounts under ESMA rules. The calculation uses a fixed lot. If the grid increases the lot as it deepens, which many do, the loss grows far faster than the square. That version is a martingale, and we cover it in the next guide.

What makes a grid survivable

Grid EAs that last tend to have a few things in common. None of them removes the risk; all of them put a ceiling on it.

1. A hard limit that fires. A maximum number of levels, a maximum total lot, or a maximum floating drawdown as a percentage of equity at which the EA closes everything and stops. The limit must be low enough to be survivable and the EA must actually honour it.
2. Capital sized to the deepest basket, not to the average. Work out the loss and the margin at the maximum number of levels. The account must survive that, with room to spare, at the leverage you actually have.
3. Fixed or slowly growing lots. Lot multipliers turn a bad grid into a disaster. A multiplier of 1.0 to 1.2 behaves very differently from 2.0.
4. A trend or volatility filter. A rule that stops adding levels, or stops opening new baskets, when ADX or ATR says the market is trending hard. It reduces how often the grid is caught.
5. A news filter. Gaps and spikes jump over levels and over stops. See our guide on the economic calendar and news filters.
6. A swap and weekend plan. Baskets held for days pay swap and carry gap risk; see weekend gaps and rollover.

Note that the first two controls are your responsibility as much as the vendor's. An EA with a sensible maximum-drawdown input is only as safe as the value you leave in it, and the default is often set to show the strategy in a good light. Set it yourself, test what happens when it fires and know the loss in dollars before you start.

How to judge a grid EA from the outside

• Ask for the deepest basket. In positions, lots and equity drawdown, on the version you are buying, and when it happened. A vendor who cannot answer does not know or will not say.
• Compare balance drawdown and equity drawdown. A grid can show a small drawdown on closed trades while sitting deep underwater on open ones. On one vendor's long-running signal we found a 13.9 percent balance drawdown beside a 72.4 percent equity drawdown; see our Waka Waka verdict.
• Read the average win against the worst loss. A grid with an average win of $10 and a worst trade of $540 needs 54 winners to pay for one bad basket.
• Look for the long record, not the best month. A record that spans a major trend, with the deepest basket shown, is worth more than ten months of calm.
• Check what the vendor changed after a bad stretch. A version history that quietly changes the limits after each deep drawdown is a warning.
• Respect our scoring rule. In our EA verdicts a grid or martingale cannot score better than High risk, however good the record, because the worst case is open-ended by design.

If a grid is what you want, treat the vendor's win rate as a measure of how often the strategy is not yet wrong. For a longer, worked look at a recovery system with a defined rule, see our zone recovery guide, and for what to do when a basket is deep, see what to do when an EA is in drawdown. Our strategy comparison puts grids next to trend and scalping systems, and our grid category lists the grid EAs in the shop with the design stated on each page.