Build your trading hours from your own trade log
Posted: Sat Oct 31, 2026 8:44 am
Copying somebody else's "best times to scalp" is a poor substitute for the data already sitting in the journal. Hours that suit one strategy, one pair and one broker can be a net drain for another.
Start by converting every trade timestamp to one timezone, preferably UTC. Group trades by hour, or by 30-minute block if there are enough, and compute net expectancy after spread and slippage for each bucket. Ignore win rate here. A bucket can win often and still lose money after costs.
The catch is sample size. With 8 trades in a bucket, the result is mostly noise. A rough floor of 30 trades per bucket is more sensible, and even that leaves wide error bars. Where the data is thin, combine neighbouring hours into broader windows before drawing conclusions.
A typical result looks like this: 07:00 to 09:30 UTC positive, midday flat to negative, and a second positive stretch when the US session overlaps. Then the schedule writes itself, with two or three defined windows and the rest of the day left alone.
Re-run the analysis every quarter. Liquidity patterns and strategy behaviour both drift.
What is the smallest sample per time bucket you would trust before dropping a session?
Start by converting every trade timestamp to one timezone, preferably UTC. Group trades by hour, or by 30-minute block if there are enough, and compute net expectancy after spread and slippage for each bucket. Ignore win rate here. A bucket can win often and still lose money after costs.
The catch is sample size. With 8 trades in a bucket, the result is mostly noise. A rough floor of 30 trades per bucket is more sensible, and even that leaves wide error bars. Where the data is thin, combine neighbouring hours into broader windows before drawing conclusions.
A typical result looks like this: 07:00 to 09:30 UTC positive, midday flat to negative, and a second positive stretch when the US session overlaps. Then the schedule writes itself, with two or three defined windows and the rest of the day left alone.
Re-run the analysis every quarter. Liquidity patterns and strategy behaviour both drift.
What is the smallest sample per time bucket you would trust before dropping a session?