Your aggregate result might be one or two hours of the day
Posted: Sat Oct 31, 2026 8:22 am
Slice the trade list by hour of the day. A very common finding is that most of the net profit comes from one or two hours, usually the London open or the overlap, while the rest of the day is flat or slightly negative after costs. That is not a flaw, it is a finding. But it means headline numbers like win rate and profit factor describe a blend of very different conditions.
It also changes the risk. The profitable window is where liquidity and volatility are best, so a change in broker pricing, a clock shift or a quieter market during that window hits the whole system at once, not just a slice of it.
The tempting move is to trade only the best hours. That can make sense, but choosing hours after seeing results is another form of optimisation. Two checks keep it honest: the hours should have a structural reason, like liquidity or a known session opening, and the effect should appear in both halves of the sample.
Also look at the weak hours separately. Sometimes they lose only because of wider spreads, and a spread filter fixes them, while sometimes the logic simply does not work there.
Which hours carry your results, and do they hold up when the sample is split in half?
It also changes the risk. The profitable window is where liquidity and volatility are best, so a change in broker pricing, a clock shift or a quieter market during that window hits the whole system at once, not just a slice of it.
The tempting move is to trade only the best hours. That can make sense, but choosing hours after seeing results is another form of optimisation. Two checks keep it honest: the hours should have a structural reason, like liquidity or a known session opening, and the effect should appear in both halves of the sample.
Also look at the weak hours separately. Sometimes they lose only because of wider spreads, and a spread filter fixes them, while sometimes the logic simply does not work there.
Which hours carry your results, and do they hold up when the sample is split in half?