Grid Trading Explained: How It Earns, When It Blows Up, and How to Set the Parameters

Grid trading has an obvious appeal: no directional call required, price oscillates inside a range, and every level is a small profit. In a ranging market it really does produce a smooth equity curve — the problem is that a single trending move gives the whole curve back at once.

How a grid earns: a 20-pip example

Setup: EUR/USD, a buy order every 20 pips, 0.01 lots per level ($0.10 per pip), 10 levels. Price falls from 1.1000 and fills one level every 20 pips.

If price oscillates between 1.1000 and 1.0800 five round trips, each level is bought and sold once for 20 pips: 20 × 0.10 = $2 per level. Ten levels × five cycles = $100.

That looks good. But note the precondition: price always returns into the range. A grid is long the assumption that it comes back, not long a direction.

Blow-up maths: fixed lots versus martingale

Fixed lots (0.01 at every level). Price falls from 1.1000 to 1.0600 — 400 pips — and all ten levels fill. Floating loss equals the sum of each level's distance to the current price, times value per pip:

400 + 380 + 360 + … + 220 = 3,100 pips → 3,100 × 0.10 = $310. On a $10,000 account that is 3.1%, entirely survivable.

Martingale (lot size doubles: 0.01, 0.02, 0.04 … 0.51, for 10.23 lots cumulative at level ten). The same 400-pip fall produces a floating loss above $15,000 — the account is gone.

That is the actual mechanism behind "grid blow-ups": not a wrong directional call, but level count × size growth × one-way distance multiplied together.

Three parameters decide survival

• Spacing: too tight and you exhaust levels and sit fully loaded; too wide and a range produces almost no fills
• Size progression: fixed, stepped or doubling. Doubling is martingale, which is mathematically certain to blow up — only the timing is unknown
• Level count and total margin: ten fixed 0.01 levels use roughly $220 of margin at 1:500; ten martingale levels need over $2,200

Margin and liquidation: leverage sets your clock

One standard lot of EUR/USD is about $108,000 notional. At 1:500 that is roughly $216 of margin; at 1:100 it is about $1,080.

For the same ten martingale levels (10.23 lots), 1:500 needs about $2,210 of margin and 1:100 needs about $11,050. Five times less leverage means five times fewer usable levels — which is exactly why a grid parameter set blows up after you move to a more tightly regulated account.

Setting spacing from ATR

A fixed pip spacing ignores volatility. With EUR/USD averaging 60 pips a day, a 20-pip grid is reasonable; at 120 pips a day, that same 20-pip grid loads all ten levels before lunch.

A more stable approach is to derive spacing from ATR: spacing = 0.3–0.5 × ATR(14). Double the volatility and the spacing widens automatically, keeping the rate of level consumption roughly constant.

Also set a total exposure cap: the notional value of all grid levels should not exceed 3–5× account equity. Above that, one trending move is enough to empty the account.

Range or trend: three objective tests

A grid only works in a range, so judging "are we ranging right now" matters more than the parameters. Three measurable tests:

One: ADX(14). Below 20 is generally non-trending; above 25 means a trend is in place, and running a grid then is adding to a losing direction.

Two: range width. Is the high-low spread over the last 20 sessions stable inside a band? A persistently widening band means a trend is forming.

Three: moving-average alignment. Are the 20, 50 and 200 MAs tangled? Tangled means range; fanned and pointing the same way means trend.

You do not need all three. But if two point to trend, either switch the grid off or halve the level count. It is not precision science, but it beats "it has felt rangebound lately".

What the equity curve looks like

Understand the shape of what you are buying. A grid curve is a long, slow climb followed by a cliff: steady small gains week after week, which reads as remarkably stable, until one directional move returns all the profit and the principal with it.

That distribution is psychologically brutal: months of positive returns encourage you to keep adding capital, and the risk peaks exactly when your exposure is largest. This is why grid blow-ups tend to arrive right after the smoothest stretch.

There is only one defence: make the tail explicit. Calculate in advance what a move of X pips costs you, write that number into the strategy description, and do not wait for it to happen.

When you must stop

• Price breaks the lower boundary of the grid and does not reclaim it within N candles
• Floating loss reaches 15% of the account, regardless of how many levels remain unfilled
• Ahead of major data or a central bank meeting — a gap will pass straight through every level
• Realised volatility jumps above twice the 20-day average

FAQ

Can a grid make money long term? In a genuine range, yes. In a trending market it gives back everything earned before. Its return distribution is many small wins plus one large loss, so the tail has to be cut off with a hard stop.

Can signals or AI switch the grid on and off? They can improve it, but "is this a range or a trend" is itself an unstable judgement, and any automatic switch lags.

To get margin and liquidation right first, read "Leverage, margin and liquidation"; to judge range boundaries, read "Price structure".