The most seductive equity curve in algo trading
No EA type sells as easily as a martingale or grid system. The equity curve rises almost linearly, drawdown looks minimal, the profit factor looks spectacular. That very beauty is the problem: it hides a risk that never appears in the statistics of the calm months, only in the one month that erases everything.
This guide explains the mechanics honestly — not to demonise these EAs, but so you understand what you are really buying when the curve looks too good. For the matching evaluation framework, see What makes a good forex EA?.
How a grid system works
A grid EA places orders at fixed intervals (e.g. every 20 pips) above and below a starting price. As the market moves, positions open one after another. When price oscillates inside the grid, the EA closes partial positions for small gains.
- Strength: in sideways markets the grid harvests a steady stream of small profits.
- Weakness: in a strong, sustained trend more and more positions open against the move, the unrealised loss grows, and there is no natural exit.
This is what the mechanics look like – on a deliberately favourable, simulated path in which every basket comes back. Watch the bracket on the right: with equal-sized levels, break-even sits exactly halfway back.
Strategy · simulation
Grid
Equal positions at fixed steps. Halfway back is already break-even – as long as price comes back.
How it works
- The EA buys and lays a grid below: one more position every 20 pips, always 0.10 lots.
- All positions share one target just above the average price.
- Because they are all the same size, the average sits exactly in the middle: halfway back is break-even, and 5 pips above that the whole basket closes in profit.
- What the animation does not show: if price keeps running against the basket, the floating loss grows with every level – without a stop there is no built-in exit.
What we check
What we check on real grid EAs: the maximum floating loss on the live account, not the closed drawdown – and whether there is a hard stop for the whole basket.
Demonstration values: a simulated EURUSD path with 0.10 lots and $10,000 starting capital, deliberately favourable – every basket and every trend works out here. Not a real trading strategy, not a backtest and not investment advice.
How martingale works
Martingale comes from gambling: after every loss the stake is doubled, so that a single win recovers all prior losses plus a small profit.
In trading that means: when a trade goes negative, the EA opens a larger counter-position (often 1.5x to 2x the lot size). When the market reverts, the whole sequence closes in profit. Many commercial EAs combine grid and martingale — the grid staggers the entries, the lot progression enlarges the positions.
The same simulated path as above, only with a ×2 lot multiplier: break-even moves in to 29% of the way back, but after just three levels seven times the starting size is in the market. An overview of all five strategy types is in EA strategies explained.
Strategy · simulation
Martingale
Each new position twice the size. A small bounce closes the basket – paid for with volume that explodes at every level.
How it works
- As with the grid, a position is added every 20 pips – but with a ×2 multiplier: 0.10 → 0.20 → 0.40 → 0.80 lots.
- The big positions drag the average down, close to the lowest entry.
- That is why a small bounce is already enough to profit. The catch: volume doubles at every level and the floating loss grows even faster – one level too many, and a single basket wipes out the gains of many before it.
What we check
What we check on real martingale EAs: the multiplier, the maximum number of levels and the volume at the last level relative to the account. A smooth equity curve is not a quality signal here.
Demonstration values: a simulated EURUSD path with 0.10 lots and $10,000 starting capital, deliberately favourable – every basket and every trend works out here. Not a real trading strategy, not a backtest and not investment advice.
Why the maths works against you
The core problem is exponential position growth against linear capital.
With 2x progression, lot size develops like this:
| Level | Lots | Cumulative |
|---|---|---|
| 1 | 0.1 | 0.1 |
| 2 | 0.2 | 0.3 |
| 3 | 0.4 | 0.7 |
| 4 | 0.8 | 1.5 |
| 5 | 1.6 | 3.1 |
| 6 | 3.2 | 6.3 |
| 7 | 6.4 | 12.7 |
After just seven consecutive losses you are trading 64x the starting size. A longer losing streak — entirely normal in a strong trend or a news gap — exceeds any real account. The market does not have to go "wrong"; it only has to run one direction longer than your capital can withstand.
This is exactly why the equity curve deceives: it shows the many times the market reverted, not the one time it did not.
What the live accounts show
Theory is one thing. Our EA review hub lists 118 commercial Expert Advisors, and for 51 of them we track a live account. 57 EAs trade a grid and 17 use martingale progression (both can apply to the same EA). The floating-loss picture, calculated straight from the dashboard:
| Metric | Grid / martingale EAs | EAs without grid |
|---|---|---|
| EAs with a published max floating loss | 9 | 11 |
| Median max floating loss | 42.7 % | 39.7 % |
| Accounts with ≥ 30 % floating loss | 7 of 9 | 8 of 11 |
| Worst value | 87.2 % (Gold Matrix EA) | 77.8 % (Zoomini) |
Read the table with two caveats. First, the gap between the groups is smaller than many expect: EAs without a grid also reach deep floating losses, for example when they hold losing positions for a long time without a stop. Second, several of the deepest grid values belong to accounts that are in profit since launch. That is exactly the mechanism from the previous chapter: the account was half gone in between, and the equity curve barely shows it.
From our own EA park: an account running a Happy Forex grid EA stood at 56 % floating drawdown on 23 July 2026, with $770 of open loss on $1,370 of equity. It recovered. The next trend in the wrong direction would not have let it. How we monitor such accounts is covered in Trading metrics: drawdown & profit factor.
Why drawdown looks harmless in the backtest
In a backtest, maximum drawdown is often measured only on a closed trade — not on the open, floating position. A grid carrying a huge unrealised loss that just barely turns positive at the end shows up as a profitable trade with low "closed" drawdown. The true moment of risk — the floating drawdown just before margin call — often does not appear cleanly in many reports.
So look at the maximum floating drawdown and the margin utilisation, not just the realised drawdown.
When grid/martingale is ever viable
There are narrowly defined scenarios in which experienced traders deliberately use this mechanic:
- Only with a real capital stop: an absolute equity drawdown at which the entire grid is closed — at a loss if necessary. This converts the unbounded tail risk into a defined maximum loss.
- Limited number of levels: at most 3–5 averaging steps, then a hard stop.
- A very small share of capital: treat the EA as a speculative position whose total loss is planned for.
- Only on pairs with pronounced mean-reverting behaviour and a low tendency to trend.
Anyone who does not accept these controls is not trading a system but a time bomb with an unknown fuse.
Non-negotiable risk controls
- A hard equity stop for the whole sequence (not per trade).
- A fixed cap on the number of open positions.
- Margin-level monitoring with an automatic emergency exit.
- No exposure over news — a news filter is not optional here.
- Never run several martingale EAs in parallel on correlated pairs: in a crash they blow the account together. The EA portfolio management guide explains why combined drawdowns are especially dangerous here.
Conclusion
Grid and martingale EAs are not magic money machines but a deliberate trade-off: many small, almost-certain gains in exchange for a rare, catastrophic loss event. Anyone who understands this and works with a hard capital stop, limited levels and a small stake can use them as calculated speculation. Anyone who falls for the smooth equity curve and plays "set and forget" will, in all probability, eventually lose everything. When in doubt: an EA with visible, controlled drawdown is safer over the long run than one that supposedly never loses.