Maximum Drawdown: How to Calculate and Interpret It
Maximum drawdown measures the largest observed decline from a previous equity peak to a later trough. It describes the path and capital pressure that a final return hides, but its meaning depends on how equity, open positions, fees, and the sampling interval are modeled.
Calculate the decline from each running peak
Track the highest equity observed up to each point. Whenever current equity is below that running peak, calculate the monetary and percentage decline. Maximum drawdown is the largest such decline, not necessarily the largest losing trade and not the difference between starting and final balance.
drawdown % = (peak equity - trough equity) / peak equity × 100The denominator is the prior peak. A drop from 12,000 to 9,000 is a 3,000 drawdown and a 25% drawdown, even if the strategy originally started at 10,000.
Distinguish closed-equity and intra-trade drawdown
A curve based only on closed trades may stay flat while an open leveraged position experiences a severe adverse move. A stronger simulation marks the open position against adverse prices and includes paid fees. Always ask whether the reported metric observes intrabar or unrealized pressure.
Audit the equity definition used at each point. Cash balance alone ignores open PnL; end-of-candle marks can miss a liquidation-threatening excursion inside the candle; marking every position at the same favorable price convention can understate portfolio pressure. For long exposure, test the adverse low; for short exposure, test the adverse high, while respecting what the data can actually establish about ordering. Include entry fees immediately and make the treatment of reserved margin, realized cash, funding, and simultaneous positions consistent across the curve.
Compare closed-only, mark-to-market, and adverse intratrade series for a sample of trades. The gap between them is itself useful: it shows how much risk disappears when only completed trades are observed. If a strategy uses a grid, DCA, or delayed stop, inspect combined open exposure rather than each leg in isolation. A final profitable close cannot retroactively make an interim margin breach executable.
Understand why recovery requires a larger percentage
Percentage loss and required recovery are asymmetric because the recovery starts from a smaller base. This is why drawdown control matters independently of expected return and why high leverage can make a strategy operationally impossible even when the final historical PnL is positive.
| Drawdown | Equity remaining | Gain required to recover |
|---|---|---|
| 10% | 90% | 11.1% |
| 20% | 80% | 25% |
| 40% | 60% | 66.7% |
| 50% | 50% | 100% |
Interpret drawdown with return, duration, and exposure
The same 15% maximum drawdown can mean different things. One strategy may recover in days with small exposure; another may remain underwater for months while using leverage and concentrated positions. Compare drawdown duration, time in market, trade concentration, liquidation distance, and the capital shared with other bots.
Separate a research rejection boundary from a later operational pause. The historical maximum is one observed sample and should not become the live stop itself. Choose a lower alert and pause threshold that leaves room for model error, latency, and an unseen sequence, then define whether the response blocks new entries, reduces exposure, or closes positions. Portfolio limits must consider correlated bots: several strategies can remain below their individual drawdown gates while creating an unacceptable account-level decline at the same time.
- Locate the peak, trough, and recovery dates on the equity curve.
- Identify open positions and market regimes during the decline.
- Check whether one symbol or grid sequence caused most of the pressure.
- Compare maximum drawdown with annualized return and Calmar ratio carefully.
- Set a live pause threshold below the historical failure boundary, not at liquidation.
Treat observed maximum drawdown as a lower-bound sample
The worst future sequence is not required to appear in the historical period. Reorder trade outcomes, widen costs, delay exits, and test hostile regimes. If the historical maximum is already close to the risk limit, there is no margin for model error or a worse sequence.
Create a drawdown stress table that changes one cause at a time: worse fill prices, an additional losing trade, correlated losses across symbols, delayed stops, and a different sequence of the same trade outcomes. Record the new depth, duration, and required recovery. The point is not to assign a probability to every hypothetical path, but to see whether modest model error breaches the account's tolerance. A strategy with no buffer should not advance at its tested size.
Recalculate one maximum drawdown
Use the equity points around the largest decline and verify the report independently.
- Find the running equity peak before the decline and the lowest later equity before recovery.
- Calculate amount and percentage drawdown using the peak as denominator.
- Inspect open-position adverse movement, fees, leverage, and liquidation distance during the trough.
- Calculate the percentage return required to recover and compare it with the strategy's typical outcome.
- Define a stricter Paper or live pause threshold with room for model error.
Drawdown interpretation checklist
Record these details before comparing strategies by drawdown.
- Peak, trough, amount, percentage, and recovery dates are identifiable.
- The equity definition includes the intended treatment of open positions and fees.
- Duration and time below the peak are reviewed beside the maximum value.
- Leverage, concentration, and liquidation risk are visible.
- The operating pause threshold leaves margin beyond the historical observation.
What maximum drawdown cannot summarize
One maximum value does not describe the frequency of losses, tail shape, liquidity risk, recovery time, or future worst case. It is sample-dependent and sensitive to equity modeling. Use it with distribution, exposure, stress tests, and explicit stop conditions rather than as a complete risk score.
Sources and further reading
Product behavior is checked against SteadyEdge documentation and code; changing exchange and platform details link to current official sources.
Trace the largest decline to its trades
Open the demo equity curve and trade journal, locate the peak and trough, and explain which positions and assumptions created the reported maximum drawdown.