Computing Expectancy From Your Own Log
The calculation itself is not the hard part. Multiply the proportion of trades that won by the average size of a win, subtract the proportion that lost multiplied by the average size of a loss, and what remains is the average outcome of taking one more trade under the same conditions. Anyone can do the arithmetic. The difficulty is that the arithmetic is only as honest as the log feeding it, and most logs are not honest by accident.
What the Log Has to Contain
Every trade needs a date, a direction, an entry, an exit, a size and a net result after costs. That is the minimum for the calculation to run at all. Useful beyond that are the planned stop, so you can tell a full loss from a partial one, and a flag for whether the trade followed the plan, which lets you compute the figure twice and compare.
Skipped sessions matter less than people expect for this particular number, since expectancy is per trade rather than per session. They matter enormously for interpreting it, because a strategy that only trades under specific conditions has an expectancy that applies only to trades taken under those conditions, and a log that does not record what was filtered out cannot tell you what the number is conditional on.
Express It in Risk, Not Currency
A figure in currency terms is unstable, because it moves whenever your position size moves, and position size moves whenever the range height moves. Two identical trades a month apart can show very different money results purely because the stop distance differed, and averaging those together produces a number that describes your sizing history rather than your method.
The alternative is to express each result as a multiple of the risk taken on that trade. A trade stopped at the planned stop counts as one unit lost. One that returned twice the planned risk counts as two units gained. The average of those multiples is comparable across sessions, across instruments and across changes in account size, and it isolates the thing you are trying to measure.
The Decisions the Calculation Forces
Several judgement calls appear the moment you sit down to compute. A trade closed manually before either the stop or the target was reached is neither a clean win nor a clean loss, and whichever bucket you put it in will shift the averages. A trade that was scratched at roughly break even sits awkwardly in a two outcome model.
There is no universally right answer, but there is a right procedure, which is to decide the treatment once, write it down, and apply it to every trade including the ones where the choice hurts. Deciding case by case is how a log drifts toward flattery, one reasonable exception at a time.
How Many Trades Before the Number Means Anything
Expectancy computed over a handful of trades is not a measurement, it is a description of those particular trades. The average win and the average loss are both averages, and averages taken from small samples move sharply when one more result arrives. A single unusually large win can carry a small sample from negative to positive, and it will carry it back on the next recalculation.
The practical sign that a sample is still too small is that the figure swings noticeably each time you add a trade. When additional trades stop moving it much, you have something worth interpreting. That point arrives later than most traders want, and there is no way to reach it faster except by continuing to trade the method the same way.
The Ways a Log Flatters
The most common distortion is omission. Trades that ended badly, particularly ones that broke the rules, are the least pleasant to record and the most likely to go missing. A log with holes always overstates the edge, and the holes cluster exactly where the information is most useful.
Second is recording gross rather than net results, which leaves the costs out of the number entirely. Third is quietly excluding a period as unrepresentative, usually a bad one. If a stretch is genuinely outside the strategy's conditions, that exclusion should follow a written rule that would also have excluded a good stretch under the same circumstances.
Computed carefully, the figure has one job. It tells you whether repeating this behaviour has a positive average result, and roughly how positive. It does not tell you what the next trade will do, and it never has, which is the part that makes it useful and the part that makes it hard to sit with.