Drawdown and recovery
How losing runs behave, what they cost, and how sizing policy decides whether an account survives one.
10 min read · intermediate · lesson 7 of 8
What this covers
- Separate the depth of a drawdown from its duration
- Work out how long a losing run a method will produce at a given strike rate
- Explain why fixed-fractional sizing slows losses without a decision
- Identify the sizing responses that convert a drawdown into a failure
- Set a drawdown limit that is a policy rather than a reaction
The two measurements
- Depth
- The fall from the account's highest balance down to its lowest, as a percentage. This is the number that determines what is required to return to flat.
- Duration
- How long the account spends below its previous high. An account can be down 4 percent for eight months, which is shallow and long.
- Peak
- The highest balance reached so far. Drawdown is measured from it, not from the starting balance, so a drawdown can begin after a run of gains.
Depth is the arithmetic problem, duration is the human one
Depth has a fixed cost, set out in the recovery table in the leverage and margin lesson: 25 percent down needs 33 percent back, 50 percent down needs 100 percent. That relationship is arithmetic and does not care how the loss happened.
Duration has no arithmetic cost at all. An account flat for a year has lost nothing but time. It is nonetheless where most methods are abandoned, because a long shallow drawdown removes the evidence that the method works while asking for it to be followed anyway.
The two failure modes are different. Depth ends an account. Duration ends a decision to keep going. A sizing policy addresses the first; knowing what a normal losing run looks like addresses the second.
How long a losing run to expect
IllustrativeAll figures are illustrative.
A method that reaches its target on 40 percent of positions loses on 60 percent. The chance of six consecutive losses is 0.6 to the sixth power, which is roughly 4.7 percent.
Over 100 positions, a run of six or more losses is not unlikely. It is close to expected.
At 1 percent risk per position, six consecutive losses is roughly 6 percent of the account. That is a normal week for that method and says nothing about whether it works.
The same calculation at 55 percent target rate gives a six-loss run a chance of roughly 0.8 percent per attempt, still ordinary across a few hundred positions.
Run this before trading a method, not during the losing run. The number is far more convincing when it is not currently being lived through.
Fixed-fractional sizing does the work without a decision
Risking a percentage rather than a fixed amount means the amount at risk falls as the account falls. Ten consecutive losses at 1 percent is not a 10 percent loss; it is closer to 9.6 percent, because each loss is 1 percent of a smaller balance.
The effect compounds in the direction that helps. A method that risks a fixed currency amount instead gets more aggressive as a share of capital exactly as results deteriorate.
The value is that no judgment is required. The rule slows the account down on the day the person operating it is least equipped to decide to slow down.
The response that turns a drawdown into a failure
Raising size to recover faster is the single most common way an ordinary losing run becomes a terminal one. It is also the most intuitive, which is why it needs a rule rather than an intention.
The arithmetic is against it in both directions. Larger size deepens the remaining losses in the run, and it does so while the evidence that the method is working is at its weakest.
The mirror image is abandoning a method mid-drawdown and adopting another. This resets the sample to zero and guarantees the next losing run is met with no idea whether it is normal, because there is no record of what normal looks like for the new method.
Setting a drawdown policy
Decide the depth that stops trading
A level at which positions stop and the method is reviewed rather than continued. Set once, in writing, away from any open position.
Decide what a review consists of
Whether the losing run is inside the range the method has produced before, or outside it. That question needs a record of prior results to answer.
Decide the reduced size on resuming
A smaller fraction of the account until the method has produced evidence again. Not a return to full size on the first position that works.
Write down the conditions for returning to full size
A number of positions, or a return to a prior balance. Anything specific enough to check an answer against.
The record is what makes any of this usable
Every judgment above needs one input: what this method has actually done. The longest prior losing run, the typical depth, the time spent below a peak.
Without that record, a drawdown is unreadable. There is no way to separate an ordinary run from a method that has stopped working, and the decision defaults to whichever feels more urgent.
Key takeaways
- Depth ends accounts, duration ends resolve. They are separate problems.
- A six-loss run is ordinary for a method that reaches target 40 percent of the time.
- Percentage risk shrinks the loss automatically as the account falls.
- Raising size during a drawdown deepens it while the evidence is weakest.
- A drawdown policy written in advance is the only kind that survives contact with one.
- None of it is readable without a record of what the method has done before.
Common mistakes
- Treating a normal losing run as proof the method has stopped working.
- Increasing size to recover faster.
- Switching methods mid-drawdown, resetting the record to nothing.
- Measuring drawdown from the starting balance rather than from the peak.
Knowledge check
OptionalRelated
- Position sizing and risk per tradeMarket Foundations
- Leverage and marginMarket Foundations