What Happens When You Close an EA's Trade by Hand

An EA is a rule set you paid for. Every time you close one of its positions, take half off, or drag a stop, you replace one of those tested rules with a rule you invented at that moment, under pressure, without a backtest. Here is what the EA actually experiences when you help it.

Most EAs track their state by magic number and by which positions are currently open. When a position with their magic number disappears, the code sees an exit and classifies it with whatever logic it has: closed at a loss, closed at a profit, or simply gone. Depending on how it was written, that can trigger an immediate re-entry because the signal is still valid, or add one to a loss-streak counter that changes the next lot size. A partial close is worse: the remaining volume no longer matches what the EA thinks it placed, and its trailing or break-even logic now acts on a size it never planned for. None of this is a bug. The code is doing exactly what it did in the backtest; you gave it inputs the backtest never contained.

A stop loss is what makes a system's risk defined. The backtest's maximum drawdown, the number you probably bought on, exists only because every losing trade was capped at the stop. Drag that stop wider, or remove it because "price will come back", and the system's risk is no longer a number. In our experience this is the single most common way a buyer turns a system with a 15% historical drawdown into an account that gets a margin call. It rarely happens on the first trade. It happens on the fifth or sixth, when the earlier four came back and taught you that it works.

Consider a system with a 35% win rate. Out of 100 trades, 65 lose, say, 1R each: minus 65R. The 35 winners must earn more than 65R between them just to break even, which means an average of roughly 1.9R per winner. But winners in a trend-following system are not evenly sized. A typical distribution is that 25 of them make 1R, 7 make 3R, and 3 make 8R or more. Those three trades contribute 24R of the total 70R. Close winners early "to lock it in" and you cap every trade at 1R or so: 35R against 65R in losses. You have converted a profitable system into a losing one without changing a single line of its code.

There is one manual action that does not corrupt the rule set: closing everything and switching the EA off during an event you decided, in advance and in writing, not to trade. The key words are in advance. Even the legitimate version should be logged: date, time, which positions were closed, at what price. Without that record you cannot later reconcile your results with the backtest, and you will not know whether the gap between them was the market or you.

· Write down the reason before you act, in one sentence. If it cannot be written, it is not a reason.

· Close the whole position, not part of it, and switch the EA off on that chart until the next clean start point.

· Never widen a stop. Tightening one is questionable; widening one is the end of defined risk.

· Record ticket, time, price, and the EA state you interrupted.

· Do it rarely. If you are doing it weekly, you do not trust the system, and the honest fix is to stop running it, not to co-pilot it.

Our three best-selling EAs: GOLD NEURON (AI) https://www.mql5.com/en/market/product/187329 · ATLAS PORTFOLIO https://www.mql5.com/en/market/product/182751