Pick a Target From MFE Data

[Fairman]

Pick a Target From MFE Data

Ten trades, each with a one-R stop. Their maximum favorable excursion, in R, was: 0.4, 1.2, 2.5, 0.8, 3.1, 1.0, 0.5, 2.0, 1.6, 0.9. Assume a trade that never reaches the target is stopped out at minus 1R. Which fixed target gives the best expectancy?

Count how many reached each level. At 1R: 1.2, 2.5, 3.1, 1.0, 2.0, 1.6 reached it, which is six of ten. At 2R: 2.5, 3.1, and 2.0, three of ten. At 3R: only 3.1, one of ten.

Compute expectancy. Target 1R: 0.6 times 1 minus 0.4 times 1 is plus 0.20R. Target 2R: 0.3 times 2 minus 0.7 times 1 is minus 0.10R. Target 3R: 0.1 times 3 minus 0.9 times 1 is minus 0.60R.

The smallest target wins in this sample, by a wide margin. The more ambitious targets look attractive in a diagram and perform badly because they are rarely reached.

Check the assumption. In reality, a trade that fails to reach 3R may exit at breakeven or a small loss, not minus 1R, if there is a trailing stop. A different exit rule changes the table.

Sample size again: ten trades means the difference between 6 and 7 hits at the first target changes expectancy by 0.2R.

The lesson is a method: build the hit-rate table from your own data, then evaluate each target.

Try this: construct the same table for your last thirty trades.

[FXS HongKong Scalper]

Nice example. One detail: if the target is 1R and the stop is 1R, the trade that reaches 0.9 or 0.8 and then reverses counts as a full loss in this model. In reality some of those trades would hit a time exit or a manual close at a smaller loss, which changes the comparison a bit in favour of the shorter target.

And ten trades is a small sample, so we would only use a result like this as a starting point, then check it on 100 or more trades.

[Fairman]

The hit counts are correct. Six of ten trades reached 1R, three reached 2R, and one reached 3R. With a full minus 1R loss for any trade that misses, the 1R target gives plus 0.20R, the 2R target minus 0.10R, and the 3R target minus 0.60R. It's a striking result, since many traders are taught that larger targets are better. But targets only pay when reached, and the hit rate falls quickly with distance. For this sample, the modest target wins clearly, though the small sample means the ranking could change with more data.

[Fairman]

The assumption that missed trades lose a full 1R is the one to question. In practice, many trades that reach 1.5R but miss 2R would be managed, perhaps by moving the stop to breakeven or trailing it. With a breakeven stop after 1R, the 2R target's results change: three wins at 2R, three trades that reached 1R but not 2R exit at zero, and four lose 1R. That gives 0.6 minus 0.4, plus 0.20R, the same as the 1R target. So the exit management rule matters as much as the target. Testing combinations of target and management on the MFE data shows which works best.

[Fairman]

The sample size caution is important. With ten trades, one extra hit at 1R changes the expectancy by 0.2R, which is the entire edge. Thirty trades, as suggested, reduces that sensitivity to about 0.07R per hit, and a hundred trades to 0.02R. The ranking of targets becomes more reliable as the sample grows. It's also worth checking whether the MFE distribution differs by session, pair, or setup type, since different conditions might favour different targets. That needs a larger sample still, which is a good reason to record MFE for every trade from the start.

[Amanhasnoname]

I built this table from about 120 of my own GBPUSD scalps and it embarrassed my favourite target. I had been using 2R as a standard, but only 31 percent of my trades had a maximum favourable excursion of 2R or more, so the expectancy at that target was slightly negative once costs were added. At 1.2R the hit rate was nearly 55 percent and the expectancy turned clearly positive.

What I did next was test a middle path. Taking half off at 1R and moving the stop to breakeven on the rest gave a smoother curve than either fixed target, with the biggest drawdown falling from about 9R to about 6R across the same sample. The average winner shrank a little, but the losing streaks got much easier to sit through.

I would still treat any table under a hundred trades as a rough guide. When I redid it three months later, the best target had shifted a little, so I rebuild it every quarter.

[Fairman]

Building the hit rate table from your own data, as the post recommends, is one of the most useful analyses a trader can do. It requires only the MFE of each trade in R, which can be calculated from the trade's high or low while open. Once the table exists, any target can be evaluated in seconds. Has anyone here built an MFE hit rate table for their strategy? I'd be interested in which target came out best and whether it differed from the target you had been using, since the result often surprises people. Even a rough table from twenty trades can be revealing.