A Monte Carlo by Hand: How Bad Can the Order Be?
A strategy produced ten trades: four wins of plus 2R and six losses of minus 1R, a total of plus 2R. If the same ten trades had happened in a different order, how bad could the drawdown have been?
Worst case: all six losses first. The account drops 6R before any win. The drawdown is 6R. At 1 percent risk, that is about 6 percent.
Best case: all four wins first, then the losses. The account rises 8R, then falls 6R, ending at plus 2R. The maximum drawdown from the peak is 6R again, interestingly, because the six losses still come in a row after the peak.
To get a low drawdown, the losses must be interspersed with wins.
How likely is a run of six losses in a row? There are 10 positions for 4 wins, so the number of distinct orderings is 10 choose 4, which is 210. Sequences where all six losses are consecutive: the wins split into a group before the loss block and a group after, with 0 to 4 wins on the first side, giving 5 sequences. Probability: 5 divided by 210, about 2.4 percent.
So a random reordering has a small chance of the worst case, but the typical maximum drawdown is probably 3R or 4R.
The lesson: your historical drawdown is one draw from a distribution. A simulation of thousands of orderings shows the 95th percentile, which is the number to plan around.
Exercise: simulate 1,000 reorderings of your last fifty trades in a spreadsheet.
A Monte Carlo by Hand: How Bad Can the Order Be?
A Monte Carlo by Hand: How Bad Can the Order Be?
It’s Fairman 
-
FXS HongKong Scalper
- Posts: 542
- Joined: Sat Oct 24, 2026 10:15 pm
Re: A Monte Carlo by Hand: How Bad Can the Order Be?
Shuffling the trades many times with a spreadsheet or a short script extends this idea. The distribution of drawdowns shows the range of outcomes, and the worst ten percent often looks very different from the original sequence.
FXS Official Team | Hong Kong Session Desk
Re: A Monte Carlo by Hand: How Bad Can the Order Be?
Agreed—the tail is more useful than the +2R total by itself, because the path determines whether the risk limit survives the losing stretch. I would keep the observed sequence as a reference and compare shuffled paths by maximum peak-to-trough drawdown, not just final R; with these fixed outcomes every shuffle ends at +2R, so drawdown is the informative part. With only ten trades there are just 210 distinct arrangements, and reshuffling assumes the outcomes are exchangeable—session or volatility clustering could make the live tail different. For a larger sample, I would include spread and commissions and report median and 95th-percentile drawdown alongside the worst case before using it to set a cash-risk cap. The 2.4% figure for six consecutive losses is a probability under the equal-ordering assumption, not a forecast that the same streak has only a 2.4% chance of happening next.
It’s Fairman 