Set a 20 pip grid on EURUSD with 0.01 lots per level and no stop. Price ticks up and down inside a range for two weeks, and the account closes thirty or forty small winners. The equity line looks like a staircase. Then the pair trends 400 pips in one direction over four sessions, and there are twenty open positions on the wrong side with an average loss of 200 pips. That is 4,000 pips of floating loss against an account that earned a few hundred.
Nothing malfunctioned. That outcome is what the design produces, and it produces it on a schedule set by how often markets trend rather than by anything the trader did wrong on the day.
What a grid actually is
A grid is a set of pending orders placed at fixed price intervals above and below the current price, each with a small take profit and, in the classic form, no stop loss. As price crosses a level, an order triggers. As it retraces by one interval, that position closes in profit and the pending order is re-armed. The system does not forecast direction. It harvests movement.
Variants change the details without changing the core. A one-directional grid only adds in one direction. A hedged grid opens positions on both sides so the net exposure stays small while individual legs are worked. A martingale grid increases the lot size at each successive level, which is a different and considerably more dangerous animal, covered in more depth in the piece on how martingale patterns are detected.
Nearly all grids run as automated code, because manually re-arming thirty pending orders across a weekend is not realistic. That means the risks are the risks of any expert advisor plus the specific ones below.
The exposure curve nobody plots
Here is the part that gets skipped. In a constant-lot grid, the number of open positions grows in proportion to how far price has travelled against you. The loss on each of those positions also grows in proportion to that distance. Multiply the two and the floating loss grows with the square of the adverse move.
A 100 pip run against a 20 pip grid gives five open positions with an average loss of 50 pips: 250 pips total. Double the run to 200 pips and you get ten positions averaging 100 pips: 1,000 pips, four times worse for twice the move. At 400 pips it is 4,000. The account does not degrade smoothly. It looks fine, then fine, then it does not exist.
Margin behaves the same way, and it is what actually ends the sequence. Each new position consumes margin while the floating loss reduces equity, so the margin level falls from both directions at once. Somewhere in that run the broker's stop out triggers and closes positions at the worst available moment. Margin calls and stop outs are the mechanism by which a grid's theoretical recovery never gets the chance to happen.
"It always comes back eventually" is true of most price series and irrelevant to a leveraged account. Recovery has to arrive before the margin runs out, and the market has no obligation to schedule its return before your stop out level.
The costs that eat the edge
Grids trade a lot, and every trade pays. Thirty small closes a week means thirty spreads paid, plus commission where the account charges it. When the take profit per level is 20 pips and the spread is 1.5 pips, roughly 7% of gross profit goes to the broker before anything else. Narrow the grid to 10 pips to trade more often and the proportion doubles.
Overnight financing is the second leak, and it is worse for grids than for ordinary strategies because the losing positions are precisely the ones held for weeks. If the direction you keep accumulating happens to be the negative-carry side of the pair, you pay swap every night on a growing stack of positions that are already underwater. Some grid systems lose more to financing than to price over a long adverse move.
Hedged grids add a third cost. Holding a long and a short in the same instrument pays the spread twice and often pays swap on both legs, and the net position is flat. You are paying carrying cost for the privilege of having no exposure.
Why prop firms screen for it
An evaluation account rewards a smooth equity curve and punishes a large drawdown, which makes grid systems look ideal until the day they do not. Firms know this, so a great many rulebooks either prohibit grids outright or cap the number of simultaneous positions per instrument. The pattern is easy to detect: orders at regular price intervals, position count rising while price moves one way, no stop loss attached to anything, lot sizes increasing after losses. Risk engines flag those signatures automatically, which is the subject of the article on grid detection in prop firms.
The consequence usually falls on the trader rather than the firm. A breach judged on pattern rather than outcome means a profitable grid can still fail an evaluation, and payout requests attract review. Anyone running one on a funded account should read the prohibited strategy clauses first and treat silence in the rules as a question to ask support, not as permission.
Where grids are defensible
I am not going to pretend the idea has no merit. Market making is a grid in professional clothing, and inventory management inside a range is a legitimate activity. The versions that survive share three properties that retail grid products almost never have.
They have a hard invalidation: a price level or a time limit at which the whole basket closes for a defined loss, accepted in advance. They size the total basket, not the individual level, so the worst case at maximum grid extension is a known percentage of the account rather than an open question. And they run on an instrument selected for its mean-reverting behaviour rather than on whatever pair the vendor's backtest happened to fit.
Even then it is a strategy with a return profile of many small gains and rare large losses, which is psychologically the hardest shape to hold. Traders tend to increase size during the long profitable stretches, so the eventual loss arrives at the largest size they ever used. If you cannot state the maximum basket loss in currency before the first order goes in, you are not running a grid, you are running an open-ended drawdown with a schedule attached. All leveraged trading carries a high risk of loss, and this structure concentrates that risk into infrequent, very large events.
"Show me the maximum basket loss in currency at full grid extension. If the answer takes more than ten seconds, the system does not have one."
— Alex Onta, Executive Director, SINGUARD
Key Takeaways
- Floating loss on a constant-lot grid grows with the square of the adverse move, not in line with it.
- Margin falls from both sides at once, so the stop out arrives before any theoretical recovery.
- High trade frequency plus long-held losers makes spread and overnight financing a large share of the cost.
- Prop firm risk systems flag the pattern itself, so profitable grids can still breach the rules.
Frequently Asked Questions
Is grid trading the same as martingale?
They are separate ideas that are often combined. A grid places orders at fixed price intervals. Martingale increases the size of each successive position after a loss. A grid with constant lot sizes grows exposure in a straight line; a grid with martingale sizing grows it geometrically, so the tenth level can be larger than the first nine combined. The martingale version is the one that removes accounts fastest.
Why do grid systems show such smooth equity curves?
Because the losing positions stay open and unrealised. Closed profits accumulate in small, regular amounts while the floating loss on the open side is not counted in the balance figure many reports display. Look at the equity line rather than the balance line, and at the maximum floating drawdown rather than the win rate. A grid can show a very high proportion of winning trades right up to the day it gives everything back.
Do prop firms allow grid trading?
Many do not, and the ones that do usually cap it. The rulebooks that prohibit it name the pattern directly, and the risk systems look for stacked opposing orders at regular intervals, position counts that grow while price moves one way, and lot sizes that increase after losses. Read the specific rules before running any such system on an evaluation, because a breach is normally judged on the pattern, not on the outcome.