Average MAE in R is each trade's MAE divided by the risk it carried. Its point lies in a boundary MAE in currency does not have: −1R is the exact place the stop should have fired, and anything beyond describes a stop that did not do its job.
Out of 100 trades
Contents
Expressing MAE in risk units reveals a marker the currency version does not have: −1R. Since a stopped-out trade is exactly −1R by definition, an MAE reaching that value means price went down to the stop, no more and no less.
That marker turns the measure into a check. An MAE of −1.4R is arithmetically impossible if the stop worked: the trade would have been closed at −1R. Each occurrence is therefore a trade where the stop was moved, removed, or crossed by an opening gap.
On winning trades, the reading changes. An MAE of −0.9R describes a trade that brushed the stop before recovering, that is a winner that nearly was a loss. Counting those says how much a performance owes to narrow margins.
MAE in R = trade MAE ÷ |entry − stop| — the value −1R corresponds exactly to the stop
A cleanly stopped-out trade has an MAE of exactly −1R. Those trades teach nothing and pull any average towards that value.
An opening gap can produce an MAE beyond −1R with no discipline failure at all. It is the only legitimate case, and it is checked against the date.
The average of this measure is of little use: it mixes stopped-out trades at −1R with winners at −0.3R. What matters is the count beyond the boundary.
| Trades beyond −1R | What it says |
|---|---|
| none | The stops fire. That is the expected case, and it is less frequent than people think. |
| a few, all on gaps | Nothing to fix. An opening gap crosses a stop without any rule being broken. |
| a few with no gap | Stops that were moved. Few in number, but each costs several times a normal loss. |
| over 10% | The stop is not a rule for you, it is an intention. That is the first job, before any setting. |
The second row is the only acceptable excuse, and it can be checked. An opening gap crosses a stop with nobody having done anything wrong, and it happens mostly at the start of the week or after a release. The trade's date is enough to settle it, and that check takes a minute per occurrence.
The rest of the table reads as a discipline job, not a setting one. A trade that went to −2.1R instead of −1R costs twice a normal loss, and no method gain compensates such a gap. That is why this count comes before optimising stop placement.
Fifteen trades out of a hundred went below a stop that should have closed them, with no gap explaining it. At −2.1R average loss, those fifteen cost 31.5R, the equivalent of thirty-one normally stopped-out trades. No method optimisation makes that back.
This is the central use, and it requires no interpretation. Filter trades whose MAE exceeds −1R, remove those falling on a gap, and count what remains. Each one is a stop moved or missing.
The point is that this check does not rest on memory. Nobody remembers pushing a stop three weeks earlier, and the arithmetic does.
On winning trades, an MAE close to −1R describes a trade that came within a hair of a loss. Many such trades mean a performance rests on very narrow margins, and that a slight deterioration in conditions would tip it over.
It is a reading of fragility rather than of result, and it appears nowhere else: those trades count as wins in every other statistic.
The first flaw is looking at the average. It mixes trades stopped at exactly −1R with winners at −0.2R, and the result lands somewhere in the middle describing neither. The useful reading is a count, not an average, and almost no tool offers it that way.
The second is forgetting gaps. A stop crossed by an opening gap produces exactly the same signature as a stop that was moved, and the accusation is unfair. The check is simple but it has to be done, otherwise this control produces guilt instead of a correction.
The third is candle resolution, inherited from MAE. An MAE computed on wide candles can show −0.95R where price actually went to −1.05R for thirty seconds. So this check understates the number of stops that should have fired.
The classic mistake. Reading the average of this measure. It adds up trades stopped at exactly −1R and winners that went through −0.2R, and lands somewhere in the middle describing neither group. The only reading that produces anything is the count of trades beyond the boundary.
| Indicator | Formula | What it measures | When to use it |
|---|---|---|---|
| Average MAE in R | MAE ÷ risk taken | The dip, with −1R as boundary | Counting the stops that did not work |
| MAE | worst point − entry | The same dip, in currency | Deciding where to place a stop |
| Trades with a stop | trades with a stop ÷ total | How many trades had one | Knowing whether the rule exists |
| Worst winner MAE | the deepest among winners | The trade that came closest | Spotting luck |
The average of this measure describes nobody. Filter trades beyond −1R, that is a list, not a number, and each line is an occurrence to look at. It is the only use producing anything.
An opening gap crosses a stop with no fault. Check the date of each occurrence: start of week, release, open after a weekend. What remains after that sort is the real figure, and it is always smaller than feared.
Optimising where to place a stop makes no sense while the stops placed do not fire. A trade at −2.1R costs two normal losses, and no placement improvement makes that back.
It is each trade's worst moment divided by the risk that trade carried. An MAE of −0.6R means price went down six tenths of the distance to your stop before turning.
That the stop did not work. −1R corresponds exactly to the stop: if price went further, the stop was moved, removed, or crossed by an opening gap. Only the last case is legitimate, and it is checked against the trade's date.
The count, almost always. The average mixes trades stopped at exactly −1R with winners that went through −0.2R, and the result describes neither. What informs is the number of trades beyond the boundary.
No. An opening gap crosses a stop with no rule broken: there was no price in between. Those occurrences are spotted by the trade's date, often a start of week or an open after a release, and they must be removed before drawing any conclusion.
That this trade came very close to a loss before leaving. Many winners in that case mean a performance rests on very narrow margins, and that a slight deterioration in conditions would tip it over. Those trades nonetheless count as victories in every other statistic.
Average MAE in R is a metric in the Reports screen, computed as soon as a trade has a known stop and available historical candles. But it is the crossing that serves, not the average: the Risk taken dimension sorts trades by risk band, and the Symbol dimension shows on which market stops get jumped.
For the fragility reading described above, the Worst winner MAE metric, in the same group, gives the extreme value directly. And the count of stops that did not work cross-checks with Trades with a stop loss, which says how many trades had a rule to enforce.
To go deeper on this: read the full guide. Every term is gathered in the glossary.