Statistics

Is this drawdown normal for this strategy? A range from its own trades

15 November 2026, 08:14
Hiroaki Yuhara
Hiroaki Yuhara
0
11

When an Expert Advisor is in a drawdown, the useful question is not "how big is it?" but "is it normal for this strategy?" A strategy with a 25% worst drawdown in its history is not in trouble at 15%, and a calm one that has never lost more than 3% is in unusual territory at 6%. This is how Halberd Portfolio Monitor 1.30 answers the question from the strategy's own closed trades, and how I checked that the calculation is right.

The method

  1. Take the closed results of one strategy (one number per closed position).
  2. Draw results at random from them, as many as the history has (at most 600), and build a new equity path. Repeat 600 times.
  3. For each path, record the deepest fall from the running peak.
  4. The 95th percentile of these deepest falls is the "normal maximum drawdown".
  5. Range = current drawdown (including the floating loss of open positions) divided by the normal maximum. Below 60% is normal, 60 to 100% is a reason to watch, 100% or more is outside the normal range.

Checking it against an exact answer

For a fair coin (+1 or -1 with equal probability), 40 trades, the distribution of the maximum drawdown can be computed exactly with a small dynamic programme: track the pair (current drawdown, maximum so far) through the 40 steps. The exact 95th percentile is 13 steps. The resampling code, fed with 20 wins and 20 losses of size 1, returned 13. I also tested the alert logic (it fires once when a strategy leaves its range, not again while it stays out, and it is armed again only after the drawdown has fallen below 80% of the range), and that the floating loss counts. All 39 automated checks of the alert and range logic passed.

Limits

  • Resampling assumes that trades are independent. Real losing streaks can be longer than the simulation says, so an alert is a reason to look closer, not a forecast.
  • At least 30 closed trades per strategy are needed; the range is only as good as that history.
  • A strategy whose behaviour has really changed will look "outside its range" for the right reason. A strategy that is simply unlucky will, by construction, do so in about 5% of the runs.

The program only reads the account and never trades. Past data only; nothing here is a forecast or advice.

#risk, expert advisor, Statistics, Drawdown, bootstrap