Drawdown: What a Losing Run Does to Judgement

The short answer

A drawdown presents you with one question — is this normal variance or is my method broken? — at the exact moment you are least equipped to answer it. The sample is small, the stakes are personal, and every rule you wrote is simultaneously up for review.

Most writing about this treats it as a test of resilience. We think that is the wrong frame. It is primarily a measurement problem, and the measurement cannot be improved from inside the drawdown. What can be improved is how much you decided before it started.

The two things the evidence does say are worth knowing, and they do not point the same way as the usual advice to stay the course:

• In the populations that have been measured, continuing did not reverse the losses. It deepened them.
• The data you would need to tell variance from breakage is almost never collected, and it is not collectable retrospectively.

What persisting looked like, where it was measured

Two studies follow people over time rather than taking a snapshot.

France, 2009–2012. The AMF collected four years of data from authorised forex and CFD providers covering 14,799 active retail investors. The share who lost money rose with the length of the window: more than 82 per cent over 2009 alone, more than 85 per cent over 2009 and 2010 together, more than 89 per cent across the whole four years. The average result moved with it — from around minus €4,989 in the first year to around minus €10,183 over two years and around minus €10,900 over four. The regulator's own summary is that clients who persevered only deepened their losses over time. Clients who placed at least 250 orders across those four years, 52 per cent of the population, averaged about minus €18,741.

Brazil, 2013–2017. Chague, De-Losso and Giovannetti observed everyone who began day trading Brazilian equity futures between 2013 and 2015. Among those who persisted for more than 300 days, 97 per cent lost money; 1.1 per cent earned more than the Brazilian minimum wage and 0.5 per cent more than a bank teller's starting salary, and the authors note that even these did so with great risk.

Neither study is about drawdowns specifically, and neither tracked whether individuals changed their approach. What they jointly undermine is the assumption buried in most advice about losing runs: that time in the market is itself the mechanism by which a drawdown reverses. In these populations, it was not.

Why you cannot read a short sequence

The core difficulty is statistical and it does not yield to reflection.

Imagine a method that genuinely wins slightly more often than it loses. Sequences of consecutive losses are not merely possible under such a method, they are expected, and they are longer than intuition suggests. A run of eight losses is unremarkable in a few hundred trades. Nothing about living through it distinguishes it from the first eight trades of a method that has stopped working.

Which means the question "is my edge gone?" cannot be answered from the drawdown. It can only be answered from a longer record, assembled before the drawdown began, and from a specification of what the method's normal behaviour looks like — written when you had no reason to flatter yourself.

The alternative, which is what most people do, is to reason from the sequence to a conclusion about the method and then to change the method. That converts a small sample problem into a permanent one, because now the pre-change record no longer describes what you are doing.

What the drawdown does to judgement

Three findings bear on the decisions people make during one.

Unrealised losses raise risk-taking. Imas showed that after a realised loss people took less risk, while an equivalent unrealised loss was followed by greater risk-taking. A drawdown is typically a mixture, and the open portion of it is the part the research flags.

Reaction intensity tracks worse outcomes. In the study of eighty day traders, those whose emotional reaction to gains and losses was more intense in both directions showed significantly worse performance. A losing run is when intensity is highest.

Self-assessment is unreliable here, and not only here. Research on confidence separates overestimation of your own performance, overplacement relative to others, and overprecision — excessive certainty in your estimate. In the experimental work, these measures were close to uncorrelated with one another, and two of the three had low reliability as measures at all. The practical reading is that introspection about how well you are doing is weak evidence in any state, and there is no reason to think it improves in a bad one. We take this apart in how to tell confidence from overconfidence.

What to decide before it happens

A drawdown is the clearest case for the if-then form of planning, because it is precisely a situation that is easy to specify in advance and hard to act on in the moment. The research on implementation intentions found their advantage appears for difficult-to-initiate actions and disappears for easy ones — which is a good description of stopping while losing.

Four things to settle in writing when nothing is at stake, as set out in what a plan has to decide in advance:

1. The limit that stops the day, and the limit that stops the week. As amounts, not feelings.
2. The drawdown at which size is reduced, and by how much, automatically.
3. The drawdown at which you stop entirely and review away from the screen.
4. What a review consists of — which numbers you will look at, decided now, so that the review cannot quietly become a search for a reason to continue.

Note that all four are about reducing exposure, not about recovering. Nothing in the evidence supports trading larger to make losses back, and the French data is a direct warning against it: the clients whose average transaction size was largest lost the most. The sizing argument is in position sizing when it is uncomfortable.

What a record makes possible

Everything above depends on having information that only exists if you collected it. The specific fields are in what to actually record in a journal, but the drawdown case makes the argument sharply: during a losing run you will want to know your longest previous losing sequence, your normal distribution of losses, and whether your recent trades met your own written criteria. All three are unavailable unless you wrote them down at the time, and all three are exactly what you need to tell variance from breakage.

A drawdown is, in that sense, an audit of the record-keeping you did months earlier. The behavioural reasons this is so hard are set out in our guide to what has actually been measured about traders.

Limits

The regulatory and academic data above describes leveraged retail populations and futures day traders over specific windows. It does not establish that any individual's drawdown has a particular cause or will or will not recover. No study we have read tested a protocol for managing drawdowns and showed that it improved outcomes.

What the evidence supports is narrower: persisting through losses did not, on average, reverse them in the populations measured, and judgement under those conditions has several documented weaknesses. Both are arguments for pre-commitment rather than for resolve.

Trading leveraged products carries a high risk of losing money quickly. Nothing on this page is advice.