Why was October worse than September? Splitting two months of made up trades four ways

[Fairman]

Why was October worse than September? Splitting two months of made up trades four ways

A journal question I find useful is not whether the month was good, but which number moved. Here are two made up months to show the split.

September: 62 trades, 45% wins, average win 1.40R, average loss 1.00R. Expectancy is 0.45 x 1.40 minus 0.55 x 1.00 = 0.08R a trade, so 4.96R for the month.

October: 41 trades, 51% wins, average win 1.20R, average loss 1.10R. Expectancy is 0.51 x 1.20 minus 0.49 x 1.10 = 0.073R a trade, giving 2.99R. The difference is minus 1.97R.

Now change one input at a time, from September toward October. Cutting the trade count from 62 to 41 costs 1.68R. Raising the win rate from 45% to 51% adds 5.90R. Shrinking the average win from 1.40R to 1.20R takes away 4.18R. Widening the average loss from 1.00R to 1.10R takes away 2.01R. The four steps add to minus 1.97R.

So the story is not that the trader got worse at picking direction. The win rate improved by six points. What changed was that winners were closed sooner and losers were allowed to run past 1R. That is a behaviour change you can look for in the trades.

One caution: the split depends on the order in which you change the inputs. Doing the average loss first, the win rate effect would come out a bit different. The total is always the same.

[Fairman]

Check how much the order matters on the same two months. Taking the changes in the reverse order, loss first, then win size, then win rate, then count, the steps are minus 3.41R, minus 5.58R, plus 8.56R and minus 1.53R, still adding up to minus 1.97R. The pattern is the same: the win rate helped, and the payoff shrank and cost more than that. The size of each piece moves, but the direction of each doesn't, so I read the split as a pointer to where to look and not as exact accounting.

[Fairman]

A good habit is to write one sentence at the end of the split naming what you'd look at in the trade log. Here: find the trades in October that were closed under 1R in profit, and the losses that ran beyond 1R, and see what happened. A loss larger than 1R usually comes from a stop that was moved or a slipped fill, and a small win from an early exit. Both are checkable on the chart, and each trade takes a minute or two. That turns a table of numbers into an actual list of 10 or 15 trades to read.

[Fairman]

Some caution on small samples again. October had 41 trades, and a win rate of 51% on 41 trades has an error of about 8 points. So the 6 point gain in win rate is within noise, and the average win and loss are noisy too. The split doesn't prove anything about cause. It does give you a way to turn a vague feeling about a bad month into specific, testable questions, such as whether wins really got cut short, which you can check on the next 40 trades. Run it on your own list and compare the two answers.

[Fairman]

A spreadsheet makes this quick. Four columns per month: trade count, win rate, average win in R, average loss in R. A fifth with expectancy using the formula win rate x average win minus (1 minus win rate) x average loss. Then a row with the total, trades x expectancy. To split, copy the September row, replace one input at a time with October's figure, and note how the total changes. It takes about five minutes a month, and you'll see after a few months which of the four numbers drives most of your swings.