A 60 trade stats pack, line by line, from one made-up log

[Fairman]

A 60 trade stats pack, line by line, from one made-up log

I generated a fake 60 trade log with a random number generator, just to show what each statistic says and how to compute it. It isn't a real account. Results are in R.

Win rate: wins divided by trades. 28 of 60, 46.7%.
Average win: +1.35R. Average loss: -0.91R.
Expectancy: win rate times average win, minus loss rate times average loss. 0.467 times 1.35 minus 0.533 times 0.91, which is +0.14R per trade.
Profit factor: gross wins divided by gross losses. 37.7R over 29.2R, 1.29.
Total: +8.5R.
Maximum drawdown: 5.0R, the deepest fall from a running peak.
Longest losing streak: 5.
Longest time below a prior peak: 26 trades, 43% of the whole log.

The line to look at is the expectancy. The standard deviation of the individual trade results is 1.18R, so the standard error of an average over 60 trades is 1.18 divided by the square root of 60, which is 0.15R. The measured +0.14R is about one standard error above zero. A profit factor of 1.29 and a positive total don't change that.

So the pack says the log is consistent with a small edge, and also consistent with none. Sixty trades is rarely enough to say which.

[FXS HongKong Scalper]

A good teaching pack, and generating a fake log is a smart way to avoid arguments about the numbers themselves. I would add the standard error of the expectancy, since with 60 trades and these averages it is likely close to the expectancy itself. That makes the point that +0.14R over 60 trades is a hint, not proof. Also worth including the longest losing streak, since it is what traders actually feel.

[Fairman]

How many trades would it take? Keep the same average of +0.14R and the same standard deviation of 1.18R. For the average to be two standard errors above zero, the standard error has to be 0.07R, which means about (1.18 divided by 0.07) squared, so around 280 trades. At 60 trades the uncertainty is 0.15R either side, so the true expectancy could plausibly be anywhere from about -0.15R to +0.45R. A trader at trade 60 can't know where in that range they stand. At trade 280 the range narrows to roughly +0.00R to +0.28R, which at least excludes a clearly losing system.

[Amanhasnoname]

This is a useful way to show it, because people argue less with made up numbers. I went through the same thing with my own log last year. After about 70 trades I had an expectancy of around 0.18R and was ready to double my size, until a friend pointed out that the spread of my results was so wide the average could easily be zero.

What I ended up doing was keeping the size flat until I had 250 trades and checking the running expectancy every 50. It bounced around a lot, dipping below zero at one point around trade 120, which would have been a horrible moment to be trading double size. By the end it settled at about 0.11R, smaller than the early number but more believable.

The time below a prior peak is the line I think new traders most need to see. Spending over 40% of the log under a previous high is normal even with an edge, and knowing that beforehand stopped me from abandoning the system halfway through.

[Fairman]

The ratio of total to drawdown is a useful single figure. +8.5R over a maximum drawdown of 5.0R is 1.7, sometimes called the recovery factor. A system that makes 8.5R and falls 5R on the way is one where the gain only modestly exceeds the pain. If the figure is 0.5, the worst drawdown was twice the total gain. It doesn't replace the others, but it tells you how well the journey matches the arrival. Comparing it across strategies in a journal helps decide where to put more size and where less, once the samples are large enough to mean something.

[Fairman]

Time under water can be as telling as depth. In this log the longest stretch below a peak was 26 trades out of 60, which means 43% of the time was spent without a new high. A trader who sees the equity curve and expects it to climb steadily will find that stretch hard, and many drop a system during it. Knowing the number in advance, and knowing that it's normal for an edge this small, sets expectations. For a scalper taking 10 trades a day, 26 trades is less than three days. For someone taking 3 a week it's two months.

[Fairman]

In a spreadsheet, with results in column A from row 2: win rate is COUNTIF(A2:A61,">0")/COUNT(A2:A61). Average win is AVERAGEIF(A2:A61,">0"). Average loss is AVERAGEIF(A2:A61,"<0"). Profit factor is SUMIF(A2:A61,">0")/-SUMIF(A2:A61,"<0"). Expectancy is AVERAGE(A2:A61). For the drawdown, use column B for the running total, column C for the running peak with MAX of the cell above and the current total, and column D for the peak minus the total. The maximum of column D is the drawdown. The standard deviation is STDEV, and dividing by the square root of the trade count gives the error.