Drawdowns look different when measured in trades

[Fairman]

Drawdowns look different when measured in trades

A scalper with a thousand trades a month will experience losing streaks that would take a swing trader years to see. A run of fifteen losers in a row, from a strategy that wins 60 percent of the time, is not impossible. It happens, and the maths is predictable.

For a 60 percent winner, the chance of a run of ten losses somewhere in a few thousand trades is far from negligible. Planning for it matters.

The emotional impact is different too. A streak of small losses over an hour feels like a collapse, even if the account is barely down.

Simulating the strategy's own statistics with random sequences gives a feel for the expected worst streak. It shows what is normal, so that a bad stretch does not trigger panic changes.

Position size should survive the expected worst streak with room to spare. A strategy that is sound but sized so aggressively that normal streaks threaten the account is a time bomb.

A rough rule is sizing so that the longest streak expected in a thousand trades would cost no more than a modest slice of the account. That slice is the one the trader can tolerate emotionally, which is a different number from the one the maths allows.

What is the longest losing streak in the log, and what does a simulation say it should be?

[Amanhasnoname]

The longest streak in a log is a fact about that sample, not a forecast. I’d estimate the win rate and payoff from a genuinely out-of-sample period, then simulate many sequences at the intended trade frequency. Report both the distribution of losing streaks and the account drawdown they imply at the planned risk per trade.

A simple shuffle assumes trades are independent, which can understate risk when losses cluster in certain sessions or volatility regimes. Resampling whole days or short blocks of trades preserves some of that clustering and gives a more conservative view. Include realistic costs and skip periods where the strategy would not actually trade.

The practical sizing question is whether the bad tail of those simulations is survivable without changing rules mid-streak. Set a maximum daily loss and a review trigger in advance, then compare live results with the simulated range. If the live sequence sits inside it, avoid treating ordinary variance as proof the edge vanished.

[FXS HongKong Scalper]

For a 60 percent winner, the chance of a run of ten losses somewhere in a few thousand trades is far from negligible.

With a 60 percent winner, the chance of ten losses in a row in a single run of ten is 0.4^10, very small, but over a thousand trades a month it becomes far more likely to happen at some point. Knowing that in advance makes a streak less of a shock.