WHAT DRAWDOWN
SHOULD YOU
ACTUALLY EXPECT?
Returns are what a strategy advertises. Drawdown is what it charges you before it pays — the peak-to-trough fall your account lives through on the way to any of those returns. It is the number that decides whether you are still in the trade when the good part arrives, and it is almost always the one quoted last, in the smallest print, if at all.
There is a mechanical reason it matters more than the headline figure: recovering from a loss is not symmetric with taking it.
| Drawdown | Gain needed just to break even |
|---|---|
| −10% | +11% |
| −25% | +33% |
| −50% | +100% |
| −80% | +400% |
A 50% drawdown does not need a 50% gain to undo — it needs a double. This arithmetic is the whole reason drawdown deserves top billing.
So the honest answer to “what should I expect?” is not a single percentage. It is a shape, and it has three parts worth separating.
1. Routine, bad-end, and the tail are different numbers
Every strategy we ship states its drawdown character in plain words rather than one figure — something like routine ~20%, occasional ~35%. Those are two different statistics: the drawdown you should expect to see fairly often, and the deeper one you should expect occasionally and not be surprised by. Neither is the worst that can happen.
Take our opt-in Aggressive Survivable strategy. Re-run over 50,000 simulated account paths, its drawdowns come out at 21% typically and 33% at the bad end — close to the “routine 20%, occasional 35%” we had originally stated. Reassuringly tidy. And measured over six months, its model prices the chance of halving the account at about one in 450.
Then reality filed its objection.
2. The model can understate the tail — ours did
The actual 2023–26 path drew down 51%, breaching the very line the model prices at one-in-450. Either that path was extraordinarily unlucky, or the simulation understates the tail. With one real sample you cannot tell — and the flattering reading is not the one to bet on.
This is a live example from our own strategy library, not a cautionary tale about somebody else’s. A simulation that says “halving is a one-in-450 event over six months” and a real three-and-a-half-year path that fell 51% cannot both be a complete description of the risk. The gap between them is the tail — the part no model measures well, because the rare events that define it are, by definition, barely present in the history you fit to.
It is also why we no longer publish that strategy’s compounding odds at all, however carefully the arithmetic was done. A number you cannot stand behind on a real sample is not a number you should put in front of someone deciding what to risk.
3. Per-trade risk is not portfolio drawdown
The most common way traders underestimate drawdown is to read the per-trade risk setting — “1% a trade, how bad can it get” — and stop there. Positions do not politely take turns.
Gap Fade fades the weekend gap at the Sunday reopen. At 1% risk per trade that sounds capped at 1%. But gaps appear across many pairs at the same reopen — typically around seven at once, sometimes sixteen — so a single bad Sunday puts many multiples of 1% at stake simultaneously. The worst weekend in the sample cost 6.1% of equity, from a strategy whose per-trade risk never changed. The profit factor and the per-trade number both looked calm; the correlation between open positions is where the drawdown actually lived.
Drawdown is spent in time, not in averages
A strategy that is profitable overall can still hand you a losing year. Swing Rider reads 1.39, 0.92, 1.39 and 1.42 across its four tested years — one losing year sitting inside a healthy overall figure. If you had started at the top of that year, the “profitable strategy” would have spent twelve months taking money from you before the average asserted itself. You live through the sequence, not the summary, and drawdown is measured along the sequence.
So, a realistic expectation
- For a diversified, out-of-sample-validated FX strategy at retail costs, plan around routine drawdowns of 15–25% and occasional ones approaching or exceeding a third. Several of ours sit there by design. A system claiming single-digit worst-case drawdown with meaningful returns is either very new, very lucky, or not measuring the tail.
- Whatever the model says the tail is, assume the real one is deeper. Ours was. Size so that the drawdown you could not quite believe would still leave you solvent and in the trade.
- Put the brake at the account level. A per-trade stop limits one position; it says nothing about ten correlated ones. The Equity Guardian halts and flattens everything at your set account-level limit — 8% below peak by default — which is the layer that actually caps a portfolio drawdown. As with any stop, gaps and slippage mean it manages the number rather than guaranteeing it.
These are backtested and simulated results, shown after out-of-sample validation. They describe how each system behaved historically. They are not a forecast, not a promise, and not a return — or a drawdown — you should expect. Live results differ, and the tail is exactly the part history reports least reliably.
See every strategy's drawdown character
A drawdown never travels alone: it sits beside a profit factor and a win rate, each of which hides a different part of the risk. And the guards that cap drawdown are not free — see what our own risk controls cost in return, measured the same way. The whole standard behind these numbers is set out on our evidence page.