Worst winner MAE is the deepest dip one of your winning trades went through before turning positive. It is an extreme value rather than an average, and that is what makes it useful: it points at the trade that came closest to disaster and still counts as a victory.
The levels of a trade
Contents
Every statistic in a journal treats a winning trade as a success. Worst winner MAE is the only one that qualifies that: among those victories, which one nearly was not? It isolates the trade that went lowest before recovering.
So the measure describes luck, and it should be called by its name. A trade entered at the wrong moment, down to within a hair of the stop, then back up to the target, appears in the journal exactly like a perfectly timed one. The result is identical, the quality of the decision is not, and no other statistic makes that distinction.
It is also a deliberately extreme value, and that is its point against average MAE. To decide where to place a stop, the average of the dips is misleading since half the trades did worse. It is the worst case that sets the bound beyond which a stop stops being workable.
Worst winner MAE = max of MAEs, among trades whose result is positive
It is most often expressed in R, because the −1R boundary immediately gives the value its meaning: it is the stop itself.
It concerns winners only. The worst MAE of a loser is −1R by construction and teaches nothing.
A value beyond −1R is arithmetically impossible if the stop worked. That winner is therefore a trade where the stop was moved, and it ended well.
| Worst winner MAE | What it says |
|---|---|
| less than −0.5R | Comfortable. None of your winners brushed the stop, and that stop may be wide for nothing. |
| −0.5 to −0.8R | Normal. The margin exists but it is not generous. |
| −0.8 to −1R | One winner brushed the stop by a hair. Tightening it would have killed it. |
| beyond −1R | Impossible if the stop worked. That winner rewards a broken rule. |
The last row is the most instructive, and it is unpleasant. A winner beyond −1R is a trade where the stop was moved and it worked. That is the worst possible combination for learning: indiscipline was rewarded, and the journal records it as a success. Spotting those trades is the only way to stop them serving as a model.
The first row is the opposite information and it is useful too. If your worst winner only went to −0.4R, none of your good trades needed the remaining 60% of your stop. Tightening it would reduce your risk per trade without cutting a single winner over the observed period, which is the rarest and most profitable case.
This +3R winner went down to −1.27R, that is beyond its own stop. It should never have existed: the stop should have closed it at −1R. It nonetheless sits among the quarter's best trades, and that is exactly the kind of trade that teaches the wrong lesson, because it rewards a rule that was broken.
The average MAE of winners indicates a tendency, the worst MAE indicates a limit. A stop placed just beyond that extreme value would have cut no winner over the observed period, which is exactly the information sought before tightening.
The caveat to keep is the sample: a quarter's worst case is not the worst possible case. A stop set exactly on that value will be crossed as soon as a slightly deeper trade appears.
A winner beyond −1R is a broken stop that ended well. Those trades corrupt learning more than any statistic can compensate, because they teach that waiting eventually pays.
Isolating and rereading them is a short and uncomfortable exercise. It is also the only one preventing an exception from becoming a habit, and that kind of habit costs more than any method setting.
The first flaw is drawing a method lesson from a case of luck. A winner that went to −0.95R was not better analysed than another: it got lucky. Many traders conclude the opposite, that their patience paid, and that conclusion ends up costing dearly.
The second is the sample. An extreme value over thirty trades is almost certainly exceeded over a hundred. Placing a stop just beyond the worst observed case gives false security, because that worst case was the worst of a short period and not of the method.
The third is that this figure does not say how many. A worst MAE of −0.9R can correspond to a single isolated trade or to fifteen almost as deep, and the stop decision is not at all the same. It reads next to the count, not in its place.
The classic mistake. Concluding that patience paid. A winner that went beyond its stop did not validate a method, it escaped a rule and ended well. That is the exact sequence turning an exception into a habit, and that habit costs more than any bad setting.
| Indicator | Formula | What it measures | When to use it |
|---|---|---|---|
| Worst winner MAE | max of MAEs, among winners | The narrowest margin of a victory | Setting the lower bound of a stop |
| Average MAE | mean of MAEs | The tendency of the dips | Placing an order of magnitude |
| MAE in R | MAE ÷ risk taken | The dip with −1R as boundary | Counting stops that did not work |
| Biggest loss | the worst realised result | A loss taken, not a dip | Spotting an outlier trade |
An extreme value points at a precise trade. Open it: what did the entry look like, was the stop moved, was there a release. Five minutes on that trade beat three hours on the group average.
The worst observed case is the worst of a sample, not the worst possible. A stop set exactly on it will be crossed by the first slightly deeper trade. Keep a margin, or accept losing that kind of trade.
An isolated extreme does not call for the same decision as fifteen almost as deep. It is the count that settles a tightening, the extreme only gives the bound.
It is the deepest dip endured by one of your winning trades before it turned positive. It is an extreme value taken among winning trades only, not an average.
Because for deciding a stop, the extreme is what counts. A stop placed at the winners' average MAE cuts about half of them, since half did worse. The worst case gives the bound beyond which no winner of the period would have been hit.
That this winning trade went beyond its own stop, which is impossible if the stop worked. So it is a trade where the stop was moved or removed, and which still ended in profit. That is the most dangerous combination for learning, because it rewards a broken rule.
With caution. The worst observed case is the worst of a sample, not the worst possible: over a hundred trades instead of thirty it will almost certainly be exceeded. A stop set exactly on it will be crossed by the first slightly deeper trade, and you must either keep a margin or accept losing those trades.
Yes, and that is precisely its use. A trade that went within a hair of the stop before recovering was not better analysed than another: it came close. Every other statistic records it as an ordinary victory, and this one is the only one making the distinction.
Worst winner MAE is a metric in the Reports screen, computed on trades whose historical candles Tradoshi could download. It reads next to average MAE in R, and that neighbourhood is the point: one gives the tendency, the other the bound.
The reading described above is completed in the Trade Review screen. Once the extreme trade is identified, its review note and its plan followed field say what the statistic never will: whether the stop was moved, and why. That is where this term stops being a figure and becomes a correction.
To go deeper on this: read the full guide. Every term is gathered in the glossary.