Grid & Martingale

Grid and martingale systems replace the usual question — *which direction will price go?* — with a different one: *can the account survive until price comes back?* A grid distributes orders across a ladder of price levels instead of committing to a single entry. A martingale increases the size of each successive entry as price moves against the open position. The two ideas are independent — a grid can use constant size, and a martingale can be triggered by signals rather than fixed spacing — but retail implementations combine them so often that they are best understood as one family.

## How a grid works

A grid is defined by a reference price, a step (the distance between levels), and a size per level. As price moves, pending orders fill at successive steps, and the basket of open positions is managed as a single unit rather than as independent trades. Each new fill pulls the average entry price toward the current price, so the basket reaches breakeven on a partial retracement rather than requiring price to travel back the full excursion.

That averaging effect is the whole mechanism, and it shapes the return distribution. Most sequences close in profit after a modest pullback, which produces a smooth-looking equity curve. The losses are concentrated in sustained directional moves that do not retrace before the level budget, the margin, or the operator's patience runs out. A grid is therefore a bet on range and mean reversion, expressed through position management rather than through an entry signal.

## Where martingale changes the risk profile

Martingale sizing multiplies the volume of each new level by a factor, commonly somewhere between 1.2 and 2.0. The consequence is structural: the distance price must retrace for the basket to break even shrinks roughly linearly with each added level, while total exposure grows geometrically. At a 2.0 multiplier, the tenth level carries 512 times the size of the first. This asymmetry is why the binding constraint in these systems is account survival and margin, not entry accuracy.

## Main variants

- **Static grid** — a fixed ladder of pending orders inside a defined range, constant size, closed at a basket target or abandoned when the range breaks. - **Directional or dynamic grid** — levels added in one direction only, often gated by a trend filter, moving average, or volatility regime. - **Hedged grid** — long and short ladders running simultaneously; behaviour depends entirely on the netting rules and on whether legs close individually or as a basket. - **Martingale averaging** — added entries with growing size, spacing driven by drawdown or signal rather than a fixed step. - **Anti-martingale / pyramiding** — size increases in the direction of profit instead of loss, inverting the return distribution. - **Scheduled cost averaging** — fixed-amount, unleveraged accumulation on a calendar. It shares the averaging arithmetic but has no recovery target and no forced exit, which places it closer to buy-and-hold than to a trading grid.

## What typically differentiates implementations

Most of the meaningful differences are not in the entry. They are in the spacing rule (fixed points versus ATR- or volatility-scaled), the sizing progression, the exit logic (basket target in currency, points or percent; trailing on the basket; partial closes), and above all the hard limits: maximum levels, maximum exposure, equity stop, time stop — or the absence of any. Instrument choice matters more than in directional systems, because spread, swap and funding accrue on baskets held for days or weeks. Whether size is fixed or scales with equity determines how a drawdown propagates.

## Common mistakes

The recurring error is reading a smooth equity curve as evidence of low risk, when the smoothness is a product of the averaging mechanism and the tail has simply not arrived yet in the sample. Closely related: reporting drawdown from closed trades only, which hides the floating drawdown that actually triggers margin calls. Others include leaving the level count or total exposure uncapped, treating a hedged grid as market-neutral when it still carries spread, swap and structural cost, optimising step and multiplier until the parameters fit one specific history, and testing on a period that happens to contain no sustained trend.

## How to evaluate and backtest a version

Start with a structural question rather than a statistical one: what is the maximum loss if every level fills and the exit never triggers? Multiply the full size ladder by the total grid span. If that number is undefined, the system has no defined risk, and no performance record changes that.

Then test against the conditions the mechanism dislikes. Include the strongest directional periods available for the instrument, not just ranging samples. Measure floating equity drawdown and peak margin used, not closed-trade statistics. Model realistic costs — spread widening, slippage, swap or funding on multi-day baskets — and check margin and stop-out levels explicitly. Randomising start dates or resampling sequences helps expose how much of the result depends on a small number of recoveries. The useful question is never how often a basket closes green, but what happens in the sequence where it does not.

The decoded versions below span the range: exchange grid bots, MT4/MT5 expert advisors with basket and hedging logic, one performance review of a live grid deployment, and a comparison of cost averaging against buy-and-hold. Each page documents that implementation's own rules and parameters.

Strategies in this concept (10)

Frequently asked questions

Is grid trading the same thing as martingale?

No. A grid describes *where* orders are placed — at fixed intervals around a reference price. Martingale describes *how much* size each successive entry uses — increasing as price moves against the position. A grid with constant lot sizes is not a martingale, and a martingale can be triggered by signals with no fixed spacing at all. They are combined frequently because the averaging effect of a grid becomes faster when sizes escalate, but they are separate design choices and carry different risk profiles.

Why do these systems show long sequences of profitable baskets followed by a large loss?

It is a property of the mechanism, not a coincidence. Averaging into a position means most sequences need only a partial retracement to close at a target, so small wins accumulate. The full excursion risk is only realised when price trends without retracing far enough, which is rarer but much larger. The return distribution is left-skewed by construction, which is why the relevant analysis is tail behaviour rather than the frequency of profitable baskets.

Does hedging make a grid safer?

It changes the risk rather than removing it. Opposing legs can neutralise directional exposure at a given moment, but the account still pays spread on both sides, accrues swap or funding, and holds margin against both legs. The unwinding logic — which leg closes first, at what target, and under what condition — is where the real exposure lives. A hedged grid that never resolves its legs can carry a locked loss indefinitely while costs continue to accumulate.

How is dollar-cost averaging different from a martingale grid?

Scheduled cost averaging buys a fixed amount at fixed intervals regardless of price, typically unleveraged and without a recovery target or forced exit. A martingale grid increases size specifically because price moved against the position, usually with leverage, a basket profit target, and a finite margin budget. They share the arithmetic of a falling average entry price, but the failure modes are different: one is exposed to long-term asset performance, the other to margin exhaustion during a single trend.

What does a credible backtest of a grid or martingale system require?

Tick-level or at least realistic intrabar data, modelled spread and slippage, swap or funding costs on positions held across sessions, and explicit margin and stop-out simulation — baskets die from margin, not from a stop-loss line. The sample must include the instrument's strongest sustained trends, and drawdown must be measured on floating equity, since closed-trade drawdown systematically understates the exposure these systems carry.

Can a grid strategy have a genuinely defined maximum risk?

Yes, if it is designed with hard caps: a maximum number of levels, a fixed size progression, a bounded grid span, and an equity or basket stop that closes everything. Multiplying the full size ladder by the total span gives the structural worst case. Systems without those caps have unbounded theoretical loss, and the practical limit becomes whatever the broker's stop-out level happens to be — which is not a risk parameter the strategy controls.

All strategy concepts · Explore all strategies