Maximum Adverse Excursion: How Far Trades Run Against You

Maximum adverse excursion is the largest drawdown a position showed against your entry while you held it. It is the only journal metric that justifies a decision about your stop rather than merely grading your outcome. Profit factor and expectancy measure what came out; this one measures the route there.

What exactly is maximum adverse excursion?

The deepest point a position reached in the wrong direction for you, measured from your entry price and bounded by the time you held it. What happens after your exit is not part of it.

That boundary matters, because it separates the metric from a related one. Maximum adverse excursion answers how much room a position would have needed. It does not answer how much profit you left behind, and the two questions get mixed up regularly.

For a trade that runs straight into profit and never falls below your entry, the figure is zero. For a stopped out trade it equals at least your stop distance, often slightly more, because the fill came in below it.

Why measure it in R?

Because the question is how the excursion relates to your risk, not how large it was in currency. A two euro drawdown is a full loss on a two euro stop distance and a fifth of one on a ten euro stop distance.

So the calculation is: excursion divided by the distance between your entry and your original stop. The result sits between zero and roughly one, with values above one marking exactly the cases where the stop was overrun.

You need that same unit for your other metrics anyway, which lets the excursion sit directly alongside the result, as covered under trading expectancy.

Where does the figure come from?

Not from your provider's statement. That records when you bought and sold but not what the price did in between. The figure therefore has to come from price data or from software that records the path of each position.

That is a concrete argument against the pure spreadsheet approach. Every other journal field can be typed in; this one cannot, because it presupposes a time series per position. Where spreadsheets otherwise run out of room is covered under trading journal spreadsheets.

A workaround to start with: record per trade only whether the price went below a marker you set, say half your stop distance. That is not a precise excursion but suffices for a first analysis and costs nothing.

How do you analyse it?

Separately for winning and losing trades. That is the entire value of this metric, and a combined analysis destroys it completely.

The excursion on your winners

This is where the real information sits. If your working trades regularly show an excursion of 0.8 R or more, they are operating right at your stop boundary, and a slightly wider stop would have saved some of the ones that did stop you out.

If the excursion on your winners instead sits mostly well below half an R, working trades need little room in your hands. Your stop is then wider than necessary, and you are paying for room you do not use.

The excursion on your losers

This distribution says less than you would expect, because it necessarily ends at your stop. The only interesting question is how many of your losers overran the stop narrowly and how many overran it substantially.

Comparing the two distributions

Only the comparison permits a decision. A wider stop saves winners with high excursions but simultaneously makes every loser more expensive, all of them. So the calculation is: how many trades would a stop wider by a given amount have saved, and what does that same amount cost across all your losing trades?

That calculation is the heart of the subject, and it comes out both ways. There are journals in which a wider stop is clearly better and others in which it makes the arithmetic worse.

Which decision follows?

One of three, and each is a deliberate change with a date attached.

1. Widen the stop and reduce position size accordingly, so risk per trade stays unchanged. That is the most common correct answer and the most commonly botched implementation, because size gets carried along rather than cut.
2. Tighten the stop where your winners need little room. That raises the number of false exits but improves the reward to risk relationship on each working trade.
3. Change nothing and improve your entries instead, where excursion is high across all trades. A uniformly high excursion points to entering too early rather than to a wrong stop.

The third option gets taken rarely and is often the most effective, because it addresses the cause rather than the symptom.

What is the obvious misapplication?

Setting the stop retroactively so that past trades look good. Placing it just beyond the largest observed excursion is not optimisation but transcribing the past.

The reason is simple: the largest observed excursion is an extreme value drawn from a sample, and the next extreme value will very probably exceed it. A stop fitted to it will be too tight again in the next stretch.

So work with the distribution rather than the maximum. The useful question is which stop distance would have covered a large share of your winners, not which would have covered all of them. And change one thing at a time, so the effect stays attributable.

What about the run in your favour?

That is the mirror metric, and it answers a different question. A position's largest favourable excursion, meaning its highest point in your favour while you held it, says nothing about your stop and something about your target.

Together the two give a complete picture of a position: how much room it needed and how much it gave back. They still belong apart, because they lead to different decisions. Excursion against you concerns the stop, excursion in your favour concerns your target and your exit.

The favourable side is most useful on losing trades. If your losers were regularly well in profit before turning, the problem sits neither in your entry nor in your stop but in your exit.

Anyone recording both should still analyse one metric at a time and change one thing at a time. Two simultaneous changes cannot be attributed afterwards.

How many trades do you need?

More than for your other metrics, because here you need a distribution rather than an average. Thirty to fifty trades under unchanged rules is the lower bound, and that applies per setup if you analyse separately.

Do not look only at the average either. A mean excursion of 0.4 R can mean every trade sat at 0.4 R, or that half sat at zero and half at 0.8 R. For your stop decision those are two entirely different situations.

How to approach analyses like this in a structured way without reading noise is covered under reviewing your trades.

What else belongs in your journal?

Four fields, three of which should be there anyway: entry price, original stop, result in R, and the excursion in R. The word original matters on the stop, because if you trail it, the first value set has to survive, otherwise the reference for the calculation is gone.

A column for execution shortfall is worth adding, because an excursion above one can have two causes: a genuine jump past the level, or a poor fill. That distinction is covered under measuring slippage.

A trading journal with price data calculates the excursion itself once entry and exit are recorded. The original stop is again the only field that has to come from your planning, and it has to be there before the trade. Which metrics carry beyond these is covered in the overview of the metrics that matter most.