Position management is where most automated systems quietly turn a decent edge into a slow bleed. The entry logic gets all the attention, but the code that runs after entry is what decides whether the equity curve survives. These are the five patterns I see most often, and what to do instead.
1. Never widen a stop to avoid a loss
It starts innocently: the trade is 10 points from the stop, and the setup "still looks valid", so the stop moves 20 points. Then 40. Then the position has no stop at all and the manager is waiting for a bounce.
The stop is not a prediction, it is a budget. If it gets hit, the trade idea was wrong and the cost was already approved. Widening it converts a series of small known costs into one unknown large one, which is exactly the shape of every blown account.
If you want more room, the place to change it is the entry parameters, tested over a full year, not the moment when a position is losing.
2. Never add to a losing position without a hard cap
Averaging down is not automatically wrong. Grid systems do it deliberately, with a defined maximum number of levels, a defined spacing, and a defined basket exit. What is wrong is averaging down without a cap, or with a cap that was chosen because "it has never gone that far".
Before any addition logic goes live, answer three questions in writing: how many additions maximum, how much total exposure at the worst case, and what closes the whole basket. If any answer is "depends", the logic is not finished.
And check the arithmetic: five additions at 1.7 multiplier is not five times the risk, it is roughly ten times the first position's risk. Run the worst case through a calculator, not through intuition.
3. Never let one trade decide the month
A single position that risks 10 percent of the account needs a 90 percent win rate to be survivable, which nobody has. The standard rule of one to two percent per trade exists because it keeps a losing streak of ten trades in the category of inconvenient instead of fatal.
If your manager computes lot size from a risk percentage, verify what the percentage is applied to: account balance, equity, or the last closed balance. Those three are different numbers after a drawdown, and using balance while the account is at equity lows quietly increases real risk exactly when you can least afford it.
4. Never trail a stop using fixed points only
A 60 point trailing distance is generous on a quiet London session and suicidal on a news spike. Trailing has to be scaled to the volatility of the instrument and the timeframe, otherwise the manager just converts trend trades into a stream of small losses.
The practical fix is ATR-based trailing with a minimum distance in points, so that the manager never trails inside the spread plus a buffer. Check what your trailing does when the position is 5 points in profit; many implementations move the stop to entry immediately, which guarantees the trade gets stopped at zero right before the move you were waiting for.
5. Never ignore correlated exposure
Opening three positions on EURUSD, GBPUSD and EURGBP looks like diversification, but those three instruments share the same euro and dollar flows. In practice it is one position at three times the size.
The same happens with five gold positions from different modules of the same EA. The manager sees five trades; the market sees one direction with five times the exposure.
Before going live, compute the worst case if all open positions lose at once and they are correlated. If that number is more than your monthly risk budget, the manager is not diversified, it is leveraged.
What to do instead
Write these as assertions in the code, not as intentions in a document: stop never moves against the position, additions capped by count and by total exposure, per-trade risk fixed and applied to the small number, trailing scaled by ATR with a spread buffer, and a correlation-aware exposure limit checked before every new order.
A position manager that refuses to open a trade because the basket is already too exposed is worth more than any additional entry signal. Entries decide how often you play; management decides whether you are still at the table next year.


