The test that removes the best trades and the worst ones
Posted: Sat Oct 31, 2026 6:48 am
Robustness checks do not have to be complicated. Two simple ones can reveal a lot.
First, remove the best ten percent of trades and see what remains. If the strategy still makes money, the edge is spread across many trades. If not, it depends on a few outliers.
Second, remove the worst ten percent. If the result becomes spectacular, then a small number of bad trades are dominating the losses, and fixing them, perhaps through a filter or stop rule, could improve things materially.
Looking at what those worst trades have in common often reveals a pattern. They might cluster at rollover, around news, or on a particular pair.
Each check takes a few minutes with a spreadsheet and can redirect the work on a strategy substantially.
It also guards against self-deception. A strategy that looks good only because of a handful of trades deserves skepticism.
What do the worst trades in your history have in common?
First, remove the best ten percent of trades and see what remains. If the strategy still makes money, the edge is spread across many trades. If not, it depends on a few outliers.
Second, remove the worst ten percent. If the result becomes spectacular, then a small number of bad trades are dominating the losses, and fixing them, perhaps through a filter or stop rule, could improve things materially.
Looking at what those worst trades have in common often reveals a pattern. They might cluster at rollover, around news, or on a particular pair.
Each check takes a few minutes with a spreadsheet and can redirect the work on a strategy substantially.
It also guards against self-deception. A strategy that looks good only because of a handful of trades deserves skepticism.
What do the worst trades in your history have in common?