How to work out whether an EA martingales before you pay for it

You do not need the source code to tell whether an expert advisor martingales. A martingale or grid system has to expose itself in three places — its input parameters, the shape of its equity curve, and the statistic the vendor chose to advertise. Any one of the three is enough to settle it.

The one question this all reduces to

The question is not whether the system loses. Every system loses. The question is whether one losing sequence has an upper bound.

With a hard stop loss on every position and a fixed lot size, the worst case is known before the trade is opened. The account can survive it, or it is sized so that it cannot, and either way the number exists.

With martingale, grid, averaging down, locking, or no stop at all, the worst case is not bounded. A long enough one-directional move takes the account to zero. The sequence ends not because the strategy contained the loss, but because the money ran out.

A system in the second group can trade profitably for years. That is what makes it sell. The equity line sits at a gentle incline while the open exposure grows underneath it, and the vendor is not lying when they publish the curve. They are publishing the part that moves.

Read the parameter list first — it is the least edited thing on the page

The input list is usually published in screenshots or in the product's set-file documentation, and it is the least edited part of a sales page. A vendor can rewrite a description twenty times. The parameters that control position sizing are harder to talk around.

Parameters that describe adding to a losing position include lot multiplier, martingale, averaging, recovery, grid step, step in pips, distance, max trades, max orders, and layers. Not all of them mean the same thing, and some are more decisive than others.

A lot multiplier above 1.0 means the next position after a loss is larger than the one before it. That is the mechanism in one number. The multiplier does not need to be dramatic. A value of 1.2 compounds the same way a value of 2.0 does, only slower.

A grid step with a maximum number of orders is bounded. The total exposure is finite, and the account-level risk can be calculated. A grid step with no maximum is not bounded. The system keeps adding positions as long as price moves against it, and the only limit is the account balance.

The absence of any stop-loss parameter is itself the answer. If every parameter concerns entry and none concerns exit, the vendor has told you what happens to losing positions without putting it in the description.

If the list is not published, ask the vendor three questions. What is the maximum number of simultaneous positions? Is there a hard stop loss on each one? What happens after several consecutive losses? If the answer to the second is no and the answer to the third is that it opens another trade, you have finished your research.

The trap in searching the description for the word itself

Searching a description for the word martingale flags the sentence “No martingale and no grid are used”, which says the opposite. It also flags every vendor who wrote that they do not average down, and those are the products most worth testing. A keyword search inverts the result.

The fix is to look for a negation in the words immediately before the term. No martingale means one thing. Martingale with nothing before it means another.

The negation check must not cross a sentence boundary. This pattern is common: “This is not a grid system. However, a martingale multiplier is applied after two consecutive losses.” Reading only the first sentence clears the product. Reading the paragraph as one string flags it for the wrong reason.

When we corrected this in our own classifier, the correction moved 57 products out of the unbounded group on the sample we were working with at the time. The error was not trivial. A check that crosses sentence boundaries misses the mechanism it exists to catch, because the vendor already wrote the disclaimer that defeats it.

What the equity curve shows and what it conceals

A strategy that closes winners and holds losers produces an equity curve that looks close to a straight line. The winning trades are realised and plotted. The losing trades are open and invisible.

The losses are unrealised, so they do not appear on the closed-trade profit line. A vendor can publish a curve with no visible drawdown and be technically accurate. The closed trades did not draw down. The open ones did.

If drawdown is measured on closed trades only, a system that never closes a loser reports a drawdown near zero while carrying an open position that could take the account. The metric is doing exactly what it was designed to do, and it was designed for a strategy that always closes.

The number to look for instead is the drawdown on the balance-and-equity curve together, or the largest open loss the account ever carried. These figures include the floating exposure that closed-trade drawdown excludes.

A curve with no visible losing periods at all is not evidence of a good system. It is evidence that losses are being held rather than closed. A system that actually closes its bad trades has red patches on the equity curve. If the red patches are absent, so is the closing.

The best-looking claim is the one to distrust

We classified 1,180 expert advisors on sale by the loss mechanism their own sales copy describes. 880 of them, 74.6%, describe a mechanism with no cap on a losing sequence.

A win rate above 90% is what a martingale grid produces by construction. It closes winners and holds losers, so nearly every closed trade is a winner. The losing trades have not closed. They are sitting open, and they do not count against the win rate until they are booked.

The claim is usually not a lie. It is an accurate description of a statistic chosen because it cannot show the risk. The vendor selected a number that is true and meaningless, and the reader filled in the meaning.

When an unbounded mechanism is still a reasonable design

A grid on a mean-reverting instrument with a small capped ladder and an account-level stop is a legitimate design. The mechanism is not the problem. The absence of a cap is the problem.

The cap is the whole difference, and a cap is a number: a maximum order count, a maximum total lot size, or a stop on total equity. If the vendor can state that number, the risk is bounded and can be sized for. The trader knows the worst case before the first trade opens.

If the vendor cannot or will not state it, the reader has their answer without needing to test anything. A vendor who has built a capped system knows the cap, because they had to type it into the parameter list. A vendor who has not built one cannot invent a number under questioning without exposing the design.

This is also why a description that mentions no mechanism is not the same as a product that has none. It is unresolved, not cleared. The absence of the word does not create a stop loss. It creates a question that still needs an answer.

Where this sits on our own product pages

Every product page on this site states the mechanism found in the vendor's original description, at the top of the page. It is the first thing a visitor reads, before the equity curve, before the settings, before the price.

Where we found none, the page says so rather than implying safety, and it says it in those terms: the description did not state a mechanism, which is not the same as the product having none. The distinction is printed on the page because the absence of a word is not evidence of the absence of a mechanism.

The classifier runs against the vendor's own text as we recorded it, not against our rewrite of it, so the verdict can be checked against what they published rather than against our summary of it. If a vendor later edits their description, our verdict does not move on its own — it moves when we re-read the page. And removing the word martingale would not make a product safer in any case. It would only make it unresolved.

We do not delist products with unbounded mechanisms. People search for them by name, and the mechanism is the most useful thing to know before paying. A reader who wants a grid system should be able to find one and see exactly what it is. What they do with that information is their decision.

The three checks above are what a description can tell you. What it cannot tell you is how the thing behaves when the market moves one way for three months, and for that there is no substitute for running it. We put the list through that window — the products that failed, and why, are on the same page as the ones that did not.