Session-based expectancy: stop blending Asia with London stats

[LondonScalper]

For about a year I ran one expectancy number for the whole account. That was a mistake.

When I finally split my trades by session, the London bucket was fine: positive after costs, stable month to month. The Asia bucket was slightly negative, with more scratches and wider relative spreads. Blended together, the account looked mediocre and I spent weeks tweaking the London playbook, which was never the problem.

How I split it now: every trade gets a session tag at entry, based on the London clock. Asia is anything before 07:00, London is 07:00 to 12:00, overlap is 12:00 to 16:00, and anything later goes in its own small bucket. Each bucket gets its own win rate, average R and cost per trade. A pivot table does the work in a couple of minutes.

The useful part is not the numbers themselves but the decisions they allow. I cut Asia to a tiny experiment budget, left London alone, and the account average improved without changing a single setup.

Separate samples, separate verdicts, same account. If your journal does not have a session column yet, add one tomorrow.

[Fairman]

Each bucket gets its own win rate, average R and cost per trade.

Good split. One thing to watch once the buckets exist is sample size per bucket, because the small ones throw around noisy numbers and it's tempting to cut them after one bad month. A rough sanity check is a confidence range on average R, using the standard deviation of R divided by the square root of the trade count. Say Asia has 40 trades with average R of -0.05 and a standard deviation of 1.1. The standard error is about 0.17, so the true number could easily be anywhere from about -0.4 to +0.3. That doesn't mean keep trading it at full size, just that the verdict is less settled than the pivot table makes it look.

[Shadow Trader]

LondonScalper, one choice in the split that changes the numbers is whether a trade belongs to the session it opened in or the one it closed in. I tag by entry like you, but about one trade in ten on my log opened before 07:00 and did most of its work after London opened, and those flattered my Asia bucket until I gave them their own tag.

The cost column is where Asia looked worst for me as well. My average EURUSD spread at entry was around 0.9 pips before 07:00 against 0.3 in London, and on a 6 pip stop that gap alone is about a tenth of an R per trade.

I would also watch the clock change weeks. If the tag is set on the London clock but the platform logs server time, the boundary shifts by an hour twice a year unless the formula handles it.

A second split that helped me was session by day of week. London on Mondays was weaker than the rest of the week, though I keep that one as a watch item since the bucket is small.

What does the overlap bucket look like for you? Mine is close to London in win rate but has a larger average loser.