A drawdown is normal when it sits inside the range your own trade history produces, in depth and in length. Your backtest defines that range and a Monte Carlo resample widens it to everything your trades could have done. Read the current drawdown against those bounds instead of against your pain level, and it stops being a guess.
Judge it against your own record, not against how it feels. Every strategy that wins over time also spends long stretches underwater, and the depth of those stretches is a property of the system: its win rate, its payoff shape, and its trade frequency set what a normal bad run looks like. A drawdown inside that range is the cost of running the strategy.
Pain is a bad ruler because it scales with recency and account size, not with statistics. The same 15R drawdown feels routine in a backtest chart and unbearable on a live account. The numbers do not change with the feeling, which is why the reading starts from the record.
Your pain threshold is not a statistic.
Four numbers from the backtest frame it. Max drawdown depth is the worst peak to trough on record. Max drawdown duration is the longest run of trades spent below a prior peak. Average drawdown duration says how long the routine dips last. Time in drawdown says what share of the whole record sat underwater. Together they describe the strategy's normal weather.
Quantprove computes all four from an uploaded trade log, and they feed the Edge Score's downside and tradability buckets directly. The practical use: write down your backtest's worst depth and longest duration. A live drawdown is unremarkable until it approaches either bound, however bad the week felt. Our demo NQ system carries a 29R worst run across 760 trades; live, that stretch would feel like a malfunction, and it sits in the record as ordinary. The key metrics guide covers where each number lives on the dashboard.
One backtest shows one ordering of your trades, and drawdown depth depends heavily on order. A Monte Carlo resample reshuffles and redraws your trades thousands of times, producing the full distribution of drawdowns the same edge could have written. The worst case in that cloud is routinely deeper than the worst case in your single backtest path.
That makes the resampled range the honest yardstick. A live drawdown deeper than the single backtest's max but inside the resampled range is the strategy exploring territory the backtest happened to skip. A drawdown beyond nearly every resampled path is a different signal entirely. Quantprove draws this distribution as the Monte Carlo simulation in Backtest analysis.
Your backtest's worst drawdown is one draw. Monte Carlo shows the full deck.
Streak math runs against intuition. A 70% win rate feels like a promise of short losing runs, yet over a few hundred trades it produces a six loss streak often enough to matter, and the drawdown around a streak is deeper than the streak itself because near misses cluster with the losses. Most traders meet their first statistically ordinary streak with the conviction that something broke.
The loss streak distribution from your own trades answers it cleanly: how long your runs get, how often each length appears, and what depth tends to surround them. A current streak inside that distribution is the system operating to spec, which is the same logic used to judge whether a prop firm challenge is survivable in the prop firm pass probability guide.
When it breaks your own bounds and keeps going. Deeper than nearly every Monte Carlo path. Longer in trades than the backtest's longest duration, then longer again. Recovery attempts that stall where they used to snap back. Each bound broken once is worth noting; bounds broken repeatedly while new trades arrive is structure, and structure in the wrong direction points past variance. One caveat: every bound here assumes the strategy you are trading is the one you backtested. Change the rules mid drawdown and the bounds reset to unknown.
Depth alone never settles it, which is why the read pairs with trend. A deep drawdown inside a healthy rolling record reads differently from the same depth at the end of a months long slide in expectancy. How to know when a strategy stops working covers that second axis, and the two reads together place the drawdown as either weather or damage.
Run the checks in order and stop at the first one that fails. Each row compares the live drawdown to a bound your own data set.
| Check | Normal | Warning |
|---|---|---|
| Depth vs backtest max | Shallower than the recorded worst | Beyond it, and still deepening |
| Depth vs Monte Carlo range | Inside the resampled distribution | Deeper than nearly all paths |
| Duration in trades | Inside the backtest's longest | Past it, with stalled recoveries |
| Loss streak length | Inside your streak distribution | New records, more than once |
| Rolling expectancy | Flat or noisy across windows | Sliding across windows |
A drawdown that passes the table is the entry fee of the strategy. One that fails several rows at once has stopped being weather. Depth also has a dial you control: risk per trade scales every drawdown up or down before the market has a say, and that sizing question is its own guide: how much should you risk per trade.