Glossary
Maximum drawdown
Last reviewed: 25 September 2026·Tradelyze
Maximum drawdown is the largest peak-to-trough decline in an equity curve over a period, expressed as a percentage of the peak. An account that rose to $110,000 and then fell to $100,000 before making a new high suffered a 9.1% drawdown. It measures the worst loss an account holder would have had to sit through.
In plain English
Maximum drawdown is the worst drop an account took from a high point before it climbed back. A small figure is not automatically good, because trading smaller positions shrinks it. At a prop firm, which funds traders who pass its paid evaluation, the size of that drop can decide whether the account survives.
New to this? Start with what backtesting is.
maximum drawdown = max of drawdown(t) across the whole period
What is maximum drawdown?
To find maximum drawdown, follow the equity curve (the value of the account over time) from start to finish. Keep track of the highest value reached so far. Each time the account falls below that high, measure the percentage fall to the lowest point before a new high. The largest of those falls is the maximum drawdown: one number describing the single worst stretch in the whole history.
Two things about that definition explain the usual misreadings. First, maximum drawdown is a maximum, not an average. One episode sets it, and the rest of the equity curve is ignored. Second, it is measured from the running peak, not the starting balance. An account that grows from $100,000 to $200,000 and then falls to $150,000 shows a 25% drawdown, even though it is still $50,000 ahead.
The most important caveat on this page
Maximum drawdown is meaningless on its own. A strategy returning 30% a year with a 10% maximum drawdown is simply better than one returning 10% a year with the same 10% drawdown. Yet the drawdown column shows the same figure for both.
Drawdown only means something next to the return that produced it and the position size used. Ranking strategies on drawdown alone rewards small positions, not better strategies.
What is an acceptable maximum drawdown?
No drawdown limit makes sense without the return figure beside it, so every number in the table below depends on context. The more useful conventions compare drawdown with annual compound return (the yearly growth rate with profits reinvested) instead of setting a fixed cap.
| Context | Limit | Source |
|---|---|---|
| Standalone system, rule of thumb | < ½ of annual compound return |
EliteTrader rule of thumb. Forum convention; no primary source. A 30% compound return tolerates a 15% drawdown on this rule. |
| Prop firm evaluation account | 5% – 10% | No primary source; a general range. Check your firm's own terms. |
| Walk-forward optimization pass criterion | < 40% | TradeStation Walk-Forward Optimizer documentation, one of its five pass criteria. |
| Read without a return figure | — | No threshold applies. Drawdown alone does not rank strategies. |
The general 5% to 10% prop firm range, which has no primary source, and TradeStation's 40% walk-forward criterion do not disagree. They answer different questions. A prop firm limit is a hard rule in the account terms, and crossing it ends the account. So the strategy's real drawdowns have to sit well inside it.
A walk-forward pass criterion is a filter on whether a strategy is worth studying further. It is applied to a figure that position sizing can scale down before any money is risked.
The practical version of the EliteTrader rule of thumb is a ratio. Divide annual compound return by maximum drawdown, and ask for a result above 2. That is the same half-of-return rule restated. The ratio says more than either number alone. It also depends much less on the position size the backtest used, because changing size moves return and drawdown together.
How much do I need to gain back to recover?
The gain required to recover from a drawdown is the drawdown divided by one minus the drawdown. It is always larger than the drawdown itself, and the gap widens sharply as the drawdown deepens.
| Drawdown | Equity left from $100,000 | Gain needed to recover |
|---|---|---|
| 5% | $95,000 | 5.3% |
| 10% | $90,000 | 11.1% |
| 20% | $80,000 | 25.0% |
| 25% | $75,000 | 33.3% |
| 33.3% | $66,700 | 49.9% |
| 40% | $60,000 | 66.7% |
| 50% | $50,000 | 100.0% |
| 60% | $40,000 | 150.0% |
| 75% | $25,000 | 300.0% |
| 90% | $10,000 | 900.0% |
| Formula | d / (1 − d) |
The gain needed is larger because it is earned on the smaller balance. A 50% drawdown leaves $50,000 of a $100,000 account, and $50,000 must double to reach $100,000 again. Each further percentage point of loss costs more to recover than the one before. That is why a deep drawdown is a different kind of problem from a shallow one, not just a bigger one.
The recovery table has a second lesson that is easy to miss: recovery takes time as well as return. Take a strategy that grows 15% a year, compounded, and suffers a one-third drawdown. It needs a 50% gain, which at its normal rate is about three years of returns (1.15 × 1.15 × 1.15 ≈ 1.52). That assumes nothing else goes wrong along the way. How long drawdowns usually last is a separate question.
How long do drawdowns last?
Drawdowns, the stretches an account spends below its last high, usually last longer than traders expect. Steadier strategies tend to have shorter ones. Here, steadier means a higher Sharpe ratio: more average return for each unit of swing in returns. The 2014 paper Two centuries of trend following (arXiv 1404.3274, §4.3) gives a rule of thumb: a typical drawdown lasts about 1 ÷ Sharpe² years. That paper's own example is a Sharpe ratio of 0.7, where typical drawdowns last about two years and four-year drawdowns are not exceptional. Before trading a strategy, ask whether you would still follow it after two years below its last high.
| Sharpe ratio | 1 / S² | Typical drawdown duration | Source |
|---|---|---|---|
| 0.5 | 4.00 | About four years | Our arithmetic, not stated in the paper. |
| 0.7 | 2.04 | About two years; four-year drawdowns are unremarkable | Stated in the paper (§4.3): “for a Sharpe of 0.7, typical drawdowns last two years while drawdowns of 4 years are not exceptional.” |
| 1.0 | 1.00 | About one year | Our arithmetic, not stated in the paper. |
| 1.5 | 0.44 | About five months | Our arithmetic, not stated in the paper. |
| 2.0 | 0.25 | About three months | Our arithmetic, not stated in the paper. |
| Rule | 1 / S² | Two centuries of trend following, arXiv 1404.3274 §4.3, states the rule and gives an example only at S = 0.7. The references it points to for more are listed in Sources. | |
The Sharpe 0.7 row matters most, because it is the only row the paper states itself. At a Sharpe ratio of 0.7, typical drawdowns run about two years, and four-year drawdowns are not exceptional, according to Two centuries of trend following. On that paper's reasoning, a long drawdown at that Sharpe ratio is not by itself evidence that the strategy has stopped working. It is what a strategy with that much return for its swings normally does.
The rule is steep. Doubling the Sharpe ratio from 0.7 to 1.4 cuts typical drawdown length by a factor of four. On the same 1 ÷ S² arithmetic, that is about two years down to about six months. That is our arithmetic, not a figure from the paper. A long drawdown tests whether you keep following a strategy, even when the drop itself is shallow.
What to do with this
Work out 1 ÷ S² from the strategy's Sharpe ratio. Decide in advance whether you would keep trading it after that long underwater, meaning below its last high. If the answer is no, the strategy does not suit you, however good the backtest looks. More optimization will not change that. In Tradelyze, use Sharpe (Daily) for this rule of thumb. On a backtest shorter than a year, no Sharpe figure is reliable enough for it.
How is maximum drawdown different from average drawdown and time underwater?
Maximum drawdown is a single worst episode. Out of an equity curve with thousands of points, it reports the one worst fall and ignores the rest. That makes it one of the least stable numbers on a backtest report. Run the same strategy on a slightly different stretch of data, and maximum drawdown can move a lot while averages move much less.
| Statistic | Data used | Stability across samples | Question it answers |
|---|---|---|---|
| Maximum drawdown | One observation — the single worst peak-to-trough episode | Low | What is the worst that happened once? |
| Average drawdown | Every drawdown episode in the curve | High | What does a bad stretch normally look like? |
| Time underwater | Every point in the curve | High | How much of the time is the account below a previous peak? |
Average drawdown is the typical depth of an account's falls from its highs. Time underwater is the share of time the account spent below a previous high. Both use the whole equity curve, which makes them far more stable across samples. Time underwater answers a question maximum drawdown cannot: not how far the account fell, but how long it stayed down. A strategy with a shallow maximum drawdown can still spend most of its life below a previous peak. Living through that feels very different from holding a strategy that recovers quickly.
None of this means maximum drawdown should be ignored. Maximum drawdown is the right number for one question: would the strategy have broken a hard limit? A prop firm's loss rule is one example. A broker's margin call, a demand for more cash when losses eat into the deposit, is another. A hard limit cares only about the single worst fall. Maximum drawdown is the wrong number for describing typical behavior, and a poor basis for ranking strategies.
Check this on your own results
Because maximum drawdown rests on one episode, treat a small difference between two strategies as noise. Say strategy A shows 12.4% and strategy B shows 13.9%. A gap that small can easily come from luck in when the losses happened. Re-run both on shifted date ranges and compare the spread of results before concluding anything. Monte Carlo simulation exists largely to show a range for this number.
What is the difference between max drawdown and trailing drawdown?
A drawdown floor is the account value you must stay above. Static maximum drawdown is measured against a fixed floor, set once from the starting balance and never moved. Trailing drawdown is measured against a floor that rises with every new equity peak and never falls back. Prop firms use both kinds, so check which one your firm applies.
Under a static rule, every profit adds room above the floor. Under a trailing rule it does not: profits permanently raise the level you must stay above.
trailing floor = highest equity so far − limit (ratchets up, never down)
Worked futures example
Take a $30,000 evaluation account with a $1,500 maximum drawdown limit. Under a static rule the floor is $28,500 for the life of the account. Under a trailing rule it starts at $28,500 and follows the peak. The numbers in this example are a constructed illustration, not measured data.
Win $300 on day one, $300 on day two and $400 on day three, taking the account to a peak of $31,000. The trailing floor rises to $29,500. Now give back $1,500 from that peak over the next five days. Equity is $29,500, the trailing floor is $29,500, and the account is disqualified — while sitting only $500 below the $30,000 it started with. The static floor of $28,500 was never approached.
The trailing rule comes in variants, and the differences matter. Some firms check the trailing floor against intraday equity in real time, including open trades. A loss that has not been closed yet can breach it. Others check it only against the end-of-day balance. That is more forgiving, because a dip that recovers before the close never counts.
Some firms stop the trailing floor rising once it reaches the account's original starting balance. Topstep's Maximum Loss Limit works this way; its help center says, “Once it reaches your starting balance, it locks permanently.” Firms differ on whether and where the floor stops, so check your firm's terms.
Trailing drawdown is explained step by step in trailing drawdown.
Tradelyze's prop firm check handles static, intraday trailing, end-of-day trailing and end-of-day balance drawdown limits. For how each variant plays out during an evaluation, see how a trailing drawdown works on the prop firm rules page. Check which variant your firm applies before reading any backtest drawdown figure against its limit.
Why is close-to-close drawdown not enough?
A close-to-close drawdown figure checks equity only at set moments, such as trade exits or session closes. It then measures the peak-to-trough fall across those snapshots. Anything that happens between two snapshots is invisible to it.
That does not match how a real-time trailing prop firm limit is checked. The real-time version of that rule is applied to intraday equity, including open positions. Picture a position that moves 2% against the account and recovers before the close. It adds nothing to a close-to-close drawdown figure, yet it can still breach a real-time trailing limit at its worst point.
So close-to-close sampling can understate the risk of a breach but never overstate it. The gap is largest for strategies that hold positions through big price swings, and for strategies with wide stops or none.
How Tradelyze measures drawdown, and which figures are a lower bound
The Max Drawdown tile and the Max DD % column in Top Trials come from the backtest engine that re-runs your script. They show the maximum drawdown that engine reports. If the engine reports none, Tradelyze rebuilds the figure from the profit and loss of closed trades, sampling equity each time a trade closes. This page does not confirm whether the engine's own figure counts dips inside trades that were still open, so do not assume it does.
MC Max DD Real→P95 on the robustness card shows two numbers. The first is the drawdown that actually happened. The second is the 95th percentile of the simulated drawdowns: the depth that 95 out of every 100 Monte Carlo simulations stayed within. Both numbers are rebuilt from your trades. Where the price data allows, Tradelyze adds each trade's lowest point, taken from the highs and lows of the bars the trade was open for. Otherwise it samples equity only at trade closes. A dip that opened and recovered inside a single bar is invisible either way, so both numbers are a lower bound. They come from a different equity path than the Max Drawdown tile, so do not compare the two directly.
The drawdown rows in each prop firm's Rule Results are rebuilt from your closed trades, in the order they closed, on that firm's drawdown type. Read them against an intraday trailing limit as a lower bound on the intraday dip, not an estimate of it. Say a strategy shows 4% against a 5% intraday trailing limit. It is not clear of that limit, because the check was not done at the level of detail the firm enforces.
Two practical responses follow. First, prefer strategies whose maximum adverse excursion (the worst open loss a trade reaches before it closes) is limited by a real stop. Then the gap between the sampled and true drawdown is limited too. Second, when a prop firm limit is close, do not treat the sampled figure as a pass. Re-check on bar-level or tick-level equity before paying for an evaluation.
Where this appears in Tradelyze
In a Tradelyze report, this is the Max Drawdown tile, and the robustness card shows a Monte Carlo worst case as MC Max DD Real→P95. To judge the whole report, not one tile, use the pre-trade checklist. Tradelyze re-runs an uploaded TradingView Pine Script strategy on the price data you upload and checks the result against your exported trade list. It then runs parameter optimization, walk-forward analysis, a four-check robustness score and prop firm rule checks. It does not place trades, give financial advice or guarantee a challenge pass, and it is in beta.
Create an account. Already a user? Open your strategies.
Going deeper
The section below goes deeper: why position size moves drawdown almost one for one, and why trading bigger does not improve the Sharpe ratio. You can skip it and still read your own report.
How does position sizing change drawdown?
Maximum drawdown grows roughly in proportion to leverage, meaning position size relative to the account. Double the risk per trade and the drawdown roughly doubles; halve it and the drawdown roughly halves. This makes maximum drawdown largely a sizing statement rather than a strategy statement.
The consequence for comparison is direct: two strategies run at different position sizes cannot be ranked on drawdown, because the difference reports the sizing choice. Any comparison has to normalize first, either by scaling both to a common volatility target or by comparing the ratio of return to drawdown, which changes far less when size changes.
What does not improve with more size is risk-adjusted return. In the idealized case the Sharpe ratio barely changes with leverage, because raising position size raises the mean return and the standard deviation of returns (how widely returns swing around their average) in roughly equal proportion. Any apparent improvement from increasing risk per trade comes from compounding effects, and those reverse past a point — beyond a certain fraction of capital risked, expected compound growth starts falling even as volatility keeps rising.
When each position is sized as a percentage of the account, so gains and losses compound, time spent in drawdown rises steadily with risk per trade. It has no turning point. In the idealized case the cause is compounding: bigger positions make deeper troughs, and each deeper trough needs a disproportionately larger gain to recover, as the recovery table shows. With a fixed number of contracts and no compounding, the picture differs. Doubling the contracts doubles every rise and every fall, so the account sits below its last high at exactly the same moments, and time underwater does not change. Either way, trading bigger buys no risk-adjusted improvement, and it buys compound growth only up to a limit.
The practical reading
Choose position size from the drawdown you can actually tolerate, then check that the resulting expected return is worth having. Doing it the other way around — choosing a return target and accepting whatever drawdown follows — is how accounts end up sized past the point where the trader can hold the position through the drawdown the strategy was always going to produce.
For a prop firm account, the same logic runs from the firm's drawdown allowance to a position size. That arithmetic is worked through in position sizing for prop firm challenges.
Stage 2 · step 7 of 18. Next in the learning path: Sharpe ratio
Frequently asked questions about maximum drawdown
What is maximum drawdown?
Maximum drawdown is the largest peak-to-trough decline in an equity curve over a period, expressed as a percentage of the peak. If an account rose to 110,000 and then fell to 100,000 before making a new high, that is a 9.1% drawdown. It measures the worst loss an account holder would have lived through.
What is an acceptable maximum drawdown?
There is no threshold that works without the return alongside it. One EliteTrader rule of thumb is to keep maximum drawdown below half of annual compound return, so a 30% compound return tolerates a 15% drawdown. Prop firm evaluation accounts are far tighter: 5% to 10% of the account is a general range with no primary source, so check your firm's own terms.
Is a low maximum drawdown always better?
No, because drawdown read alone is meaningless. A strategy returning 30% a year with a 10% drawdown is simply better than one returning 10% a year with the same 10% drawdown, and the drawdown figure is identical for both. A low drawdown usually reflects small position sizing rather than a better strategy.
How much do I need to gain back to recover from a drawdown?
The gain required is the drawdown divided by one minus the drawdown. A 20% drawdown needs a 25% gain to get back to the previous peak, a 25% drawdown needs 33.3%, a one-third drawdown needs 50% and a 50% drawdown needs 100%. The required gain grows faster and faster as the drawdown deepens.
Why does a 50% drawdown need a 100% gain to recover?
Because the gain is earned on the reduced balance. An account of 100,000 that falls 50% is left with 50,000, and 50,000 has to double to get back to 100,000. Each further percentage point of drawdown costs more in recovery than the last, which is why deep drawdowns are a different kind of problem from shallow ones, not just bigger.
How long do drawdowns last?
Drawdowns usually last longer than traders expect, and steadier strategies (higher Sharpe ratio) tend to have shorter ones. Two centuries of trend following (arXiv 1404.3274) gives a rule of thumb: a typical drawdown lasts about 1 ÷ Sharpe² years. The paper's example is a Sharpe ratio of 0.7, where typical drawdowns last about two years and four-year drawdowns are not exceptional. Before trading a strategy, ask whether you would still follow it after two years below its last high.
What is the difference between maximum drawdown and average drawdown?
Maximum drawdown is a single worst episode, so it rests on one stretch of the data and is one of the least stable numbers on a backtest report. Average drawdown and time underwater use every data point in the equity curve, which makes them far more stable across samples and far more informative about what holding the strategy actually feels like.
What is time underwater?
Time underwater is the proportion of the backtest during which equity sat below a previous peak, and the longest unbroken such stretch. Time underwater answers a different question from drawdown depth: not how far the account fell, but how long it stayed down. A strategy with a shallow maximum drawdown can still spend most of its life underwater.
What is the difference between max drawdown and trailing drawdown?
Static maximum drawdown is measured against a fixed floor set from the starting balance and it never moves. A trailing drawdown floor rises with every new equity peak and never falls back. The trailing version is never easier to satisfy, because profits you earn permanently raise the level you must stay above.
How can a trailing drawdown disqualify me when my account is barely down?
Because the floor follows your peak, not your starting balance. On a 30,000 account with a 1,500 trailing drawdown, 1,000 of winning days lifts the peak to 31,000 and the floor to 29,500. Giving back 1,500 from that peak puts you at 29,500 and breaches the limit, while your account is only 500 below where it started.
Why is close-to-close drawdown not enough?
Because prop firm trailing drawdown is often evaluated on intraday equity including open positions, while a close-to-close figure only samples equity at trade exits or session closes. Any excursion between those samples is invisible. A backtest reporting only close-to-close drawdown systematically under-reports breach risk, and the gap is largest for strategies that hold through volatility.
How does position sizing change maximum drawdown?
Almost proportionally. Maximum drawdown grows roughly in step with position size, so doubling risk per trade roughly doubles the drawdown. This makes maximum drawdown largely a sizing statement rather than a strategy statement, and it is why comparing the drawdowns of two strategies run at different position sizes tells you about the sizing, not the edge.
How does Tradelyze calculate maximum drawdown?
The Max Drawdown tile and the Max DD % column in Top Trials show the maximum drawdown reported by the backtest engine that re-runs your script. If the engine reports none, Tradelyze rebuilds it from closed-trade profit and loss. MC Max DD Real→P95 on the robustness card and the drawdown rows in prop firm Rule Results are rebuilt from your trades and can miss a dip inside an open trade, so treat them as lower bounds.
Sources
- Y. Lempérière, C. Deremble, P. Seager, M. Potters and J. P. Bouchaud, Two centuries of trend following, arXiv 1404.3274, submitted 12 April 2014, §4.3 — “the typical duration of a drawdown is given by 1/S² (in years) for a strategy of Sharpe ratio S. This means that for a Sharpe of 0.7, typical drawdowns last two years while drawdowns of 4 years are not exceptional.” The paper gives no other worked value. For more on the topic it points to J. P. Bouchaud and M. Potters, Theory of Financial Risk and Derivative Pricing (Cambridge University Press, 2003), and to P. Seager et al., The statistics of drawdowns, listed there as in preparation in 2014. This page has not checked either reference for a derivation. Retrieved 28 July 2026; the quoted passage was re-checked against the arXiv PDF on 15 September 2026.
- Rows other than S = 0.7 in the drawdown-duration table are our own arithmetic, applying 1/S² at 0.5, 1.0, 1.5 and 2.0. They are correct arithmetic but do not appear in the paper.
- TradeStation, About the Walk-Forward Optimizer, help topic
tswfo/about_wfo— the five default pass criteria, the last of which is “Has a Maximum drawdown of less than 40%.” Retrieved 14 September 2026. - EliteTrader thread
212930, Acceptable drawdown amount? — the rule of thumb that maximum drawdown should stay below half of annual compound return. Forum convention; no primary source. Retrieved 28 July 2026. - Topstep help center article
8284204, What is the Maximum Loss Limit? — “Once it reaches your starting balance, it locks permanently.” The trailing floor freezes at the original starting balance, not at the starting balance plus the profit target. Retrieved 14 September 2026. - Tradelyze implementation, reviewed 15 September 2026 — the Max Drawdown tile and the Max DD % column show the maximum drawdown reported by the backtest engine that re-runs the script, and fall back to a rebuild from closed-trade profit and loss when the engine reports none; MC Max DD Real→P95 is rebuilt from the trades, adding each trade's bar-resolution low point where it can be reconstructed and otherwise sampling equity at trade closes, so it is a lower bound; prop firm drawdown rows are rebuilt from closed trades in the order they closed and apply static, intraday trailing, end-of-day trailing and end-of-day balance drawdown limits.
- The 5% to 10% prop firm evaluation range in the acceptable-drawdown table has no primary source; it is a general range, and each firm's own terms decide the real limit.
- Recovery percentages in the recovery table are arithmetic, computed as d / (1 − d). They are not sourced statistics. The $30,000 evaluation-account walkthrough and the figure it feeds are constructed illustrations, not measured data, and follow a different path from the example on the prop firm rules page.