How long does drawdown recovery take—if your loss limit gets there first?

What does “recovering a drawdown” actually mean?

A trading account previously reached $10,000 and now holds $9,000. There are three different questions hiding inside “How long until I get back?”

1. How much must it gain? $1,000, or 11.11% of the current $9,000. That is arithmetic.
2. How long did a past recovery take? The interval between the chosen observation and the first later observation at the old peak. That is a measurement of history.
3. What is the chance of recovering within the next 100 trades, before a lower loss limit ends the attempt? That requires a model of future paths.

The drawdown recovery calculator answers the arithmetic and historical questions, with an optional constant-return scenario. This article addresses the third question. A fixed monthly return and a sequence of variable trade outcomes are different assumptions.

“Recovery time” also needs a starting point. Time underwater runs from the old peak; a trough-to-recovery interval starts later. Here the clock starts at the current $9,000 observation. Calling it the trough would assume the account cannot fall further—the very uncertainty we need to preserve.

Can a positive edge still fail to recover?

Consider a deliberately simple hypothetical account. Every trade independently earns $100 with probability 55%, or loses $100 with probability 45%. These are net outcomes; no further fee is deducted. The dollar amount stays fixed as the balance changes.

Prior peak / recovery target

$10,000

Current balance

$9,000

Fixed amount per outcome

$100 = 1R

Stop balance

$8,500

The account needs 10R of net gains to recover and has only 5R of room before stopping. It stops on the first trade close at either boundary. There is no compounding, trailing floor, daily limit, deposit, withdrawal or intratrade price path in this model.

The second line is an appealing shortcut. It tells you when the expected net gain of an unstopped sequence reaches $1,000. It does not tell you how many paths get there before first touching $8,500. A path that hits the loss limit is no longer allowed to collect its hypothetical later gains.

What happens by 20, 50, 100 or 200 trades?

Instead of averaging account balances, count the probability of each first outcome. Every path belongs to exactly one of three groups: recovered before stopping, stopped before recovering, or still between the boundaries when the horizon ends.

With unlimited trades under these same assumptions, the probability of recovering before the stop approaches 66.62%. The other 33.38% reach the stop first. This is the absorbing-boundary result, not a claim that markets eventually deliver either outcome on schedule.

Nor is “unresolved” another word for failed. At 100 trades, 9.79% of paths have reached neither boundary. They could later recover or stop. A useful report keeps that group visible instead of quietly adding it to whichever outcome makes the headline simpler.

Why can “average recovery time” look reassuring?

Among the paths that recover by trade 100 in this example, the average recovery takes 46.76 trades. It sounds much better than the 100-trade shortcut. But it describes only the 58.65% that already succeeded within that window.

The stopped 31.56% have no recovery time under the stopping rule. The unresolved 9.79% have not yet supplied one. Dropping both groups makes a fast recovery average compatible with a substantial chance of never recovering within the permitted attempt.

Even the choice of horizon changes the average. At 200 trades, slower recoveries enter the successful group and its mean rises to 56.22 trades. Nothing about the strategy improved or deteriorated; we simply watched longer.

The distinction is familiar in time-to-event analysis: an unfinished observation has a duration known only up to the observation cutoff. NIST calls this right censoring. Here a stopped account is a separate terminal outcome; treating it as merely an account we stopped observing would change the question.

How sensitive is recovery to the assumed edge?

For a model this small, the assumed edge deserves immediate scrutiny. Keep the $1,000 recovery gap, $500 remaining room, $100 fixed outcomes and 100-trade horizon unchanged. Change only the win probability:

A precise calculation is not evidence that the input is precisely known. A measured 55% win rate can be uncertain, outcomes can cluster, and the strategy can change. The exact recursion removes simulation sampling noise; it removes none of those model risks. Trade-sample uncertainty and what a win-rate summary loses about trade order deserve their own checks.

Position sizing also changes the experiment. Fixed dollar outcomes keep the distance to each boundary measured in the same R. Fixed fractional sizing changes dollar gains and losses with equity. Doubling risk is therefore not a valid way to halve the recovery date from this table; it changes boundary-hitting probabilities as well.

How can you frame a recovery scenario you can actually test?

A useful setup names the current capital, the old peak, the remaining loss allowance, the trade-outcome assumptions and the horizon. It then reports the chance of reaching the recovery target before the stop, alongside the stopped and unfinished shares. Duration summaries state which paths qualify.

In the hypothetical example, a model starting fresh at $9,000 would need a recovery goal $1,000 higher (11.11% of current capital) and a static stop $500 lower (5.56% of current capital). These percentages are rounded descriptions of the dollar boundaries, not suggested trading limits. A maximum loss measured from the original $10,000 peak has a different denominator.

PropSurvival’s personal rules simulator can model a stated goal, a static loss allowance and a horizon against supplied trade statistics. That is a natural continuation of the decision. The exact recovery curve and conditional averages in this article are a separate research calculation; they are not claimed as app outputs, and this article does not transfer its inputs into the app. Production settings and cost assumptions must be read in the tool itself.

What else does a recovery estimate need to disclose?

Can trade counts be converted into a calendar date?

Only with an additional assumption about when trades occur. At exactly five trades per active trading day, 100 trades span 20 such days by arithmetic. That says nothing about weekends, missed setups, pauses or a changing opportunity rate. More trades per day also need not preserve the same edge. This study makes no calendar forecast.

Does a long drawdown prove a strategy has stopped working?

No single duration settles that question. Rej, Seager and Bouchaud’s research on drawdown depth and length examines how both can inform assessment under a specified stochastic model. Our example instead holds the win probability fixed; it cannot distinguish bad luck from a deteriorating edge. That requires evidence beyond the numbers entered here.

What if the floor trails, or losses exceed the stop?

Then this table is the wrong model. A trailing floor requires tracking the account’s high-water mark; intratrade breaches require an intratrade path; variable losses require their own distribution. A stop order is not evidence that every realized loss equals the planned loss. The example’s fixed ±$100 trade closes are assumptions, not execution guarantees.

How were these probabilities calculated?

Define the current balance as 0R, recovery as +10R, and the stop as −5R. At the start, all probability mass is at 0R. After each trade, move 55% of every still-active state’s mass up one R and 45% down one R. Mass reaching a boundary is recorded there and never traded again.

This finite-state recursion sums all possible paths without randomly sampling them. Results are deterministic probabilities, computed in double precision; there is no simulation count or seed. The conditional mean is the sum of each recovery trade number times its first-recovery probability, divided by cumulative recovery probability at that horizon.

The unlimited-horizon recovery probability is independently checkable using the standard two-boundary random-walk formula. With r = 0.45/0.55, it is (1 − r⁵)/(1 − r¹⁵) = 0.6661876933. The finite-horizon probabilities approach that limit.

Download the dataset and full per-trade probability series (JSON). Method version: ps-drawdown-recovery-first-passage-1.0.0. All account amounts and trading parameters are illustrative; all table, figure and duration values are model-derived. This is educational research, not investment or trading advice.

Sources and evidence

1. MIT, Leighton & Rubinfeld: Random Walks, lecture notes dated 12 December 2006, §1. Mathematical basis for a random walk with two absorbing boundaries. Our account parameters and finite-horizon figures are our own calculation. Accessed 13 October 2026.
2. NIST/SEMATECH e-Handbook: Censoring, §8.1.3.1. Basis for the distinction between an observed event time and an unfinished duration at a cutoff. Accessed 13 October 2026; no page revision date stated.
3. Rej, Seager & Bouchaud: You are in a drawdown. When should you start worrying?, version 2, 21 July 2017. Context on depth and duration under a Brownian model; it is not the source of this article’s discrete-trade numbers. Accessed 13 October 2026.
4. PropSurvival methodology and personal rules simulator, inspected 13 October 2026. Product-scope references; the research model here remains separate from the production engine.