The EA that kept blocking manual trades: a case study
Posted: Sat Oct 31, 2026 6:51 pm
Here is a case that comes up in different forms. A trader runs an EA on a pair and also trades manually on the same account, on the same symbol, now and then. For weeks, the EA's results look noticeably off from its test results, and nobody can explain why.
The first suspects are the obvious ones: spread, slippage, execution speed. All get checked and none fits. Then somebody reads the EA's logic more carefully. The EA limits itself to a maximum number of open positions on the symbol, and it counts every open position, not only its own. Whenever a manual trade was open, the EA saw the limit reached and skipped valid signals. Test runs never had a manual trade in them, so the tests showed more entries than live.
That was only half of it. The twist came from the other direction. The EA also had a routine that adjusted stops on all open positions of the symbol, and it did not filter by magic number, the identifier that marks which orders belong to the EA. So manual trades had their stops quietly moved by the EA's trailing logic. Several manual trades that were meant to run got closed early by a stop that nobody had placed.
The lesson is that an EA and a human share one account, and unless the code explicitly separates their orders, they interfere in both directions. Entries get blocked, stops get moved, and closing functions that run on everything on the symbol can close manual positions too.
Checks worth doing: confirm the EA filters by magic number in every function that counts, modifies, or closes orders. Look at the position limit logic. Avoid trading the same symbol manually while an EA manages it, or move one of them to a separate account where the account type allows it. On a netting account the problem is worse, because both end up in the same single position.
Has anyone been caught out by an EA touching manual orders, and what was the clue that revealed it?
The first suspects are the obvious ones: spread, slippage, execution speed. All get checked and none fits. Then somebody reads the EA's logic more carefully. The EA limits itself to a maximum number of open positions on the symbol, and it counts every open position, not only its own. Whenever a manual trade was open, the EA saw the limit reached and skipped valid signals. Test runs never had a manual trade in them, so the tests showed more entries than live.
That was only half of it. The twist came from the other direction. The EA also had a routine that adjusted stops on all open positions of the symbol, and it did not filter by magic number, the identifier that marks which orders belong to the EA. So manual trades had their stops quietly moved by the EA's trailing logic. Several manual trades that were meant to run got closed early by a stop that nobody had placed.
The lesson is that an EA and a human share one account, and unless the code explicitly separates their orders, they interfere in both directions. Entries get blocked, stops get moved, and closing functions that run on everything on the symbol can close manual positions too.
Checks worth doing: confirm the EA filters by magic number in every function that counts, modifies, or closes orders. Look at the position limit logic. Avoid trading the same symbol manually while an EA manages it, or move one of them to a separate account where the account type allows it. On a netting account the problem is worse, because both end up in the same single position.
Has anyone been caught out by an EA touching manual orders, and what was the clue that revealed it?