What Is a Good Maximum Drawdown for a Trading Bot?

What maximum drawdown means

A drawdown begins when account equity falls below a previous peak. It ends only when equity reaches a new peak. Maximum drawdown is the deepest peak to trough decline recorded during the selected period.

If a trading account rises from 10,000 to 12,000 and then falls to 9,600 before recovering, the drawdown is measured from 12,000 rather than from the original capital. The decline is 2,400, or 20% of the peak.

Maximum drawdown answers a different question from total return. Return describes where the strategy finished. Drawdown describes the largest loss path an investor had to endure on the way there.

How maximum drawdown is calculated

At every point in the equity series, record the highest equity reached so far. Compare current equity with that running peak. The largest percentage decline is the maximum drawdown.

The basic calculation is peak equity minus trough equity, divided by peak equity. The result is normally expressed as a positive percentage describing the size of the loss.

The calculation depends on the equity data supplied to it. A daily account series, a bar by bar strategy series and a list containing only closed trades can produce different results for the same system.

• Peak equity: the highest account value reached before the decline
• Trough equity: the lowest value reached before a new peak
• Drawdown depth: the percentage loss from peak to trough
• Recovery: the return to the previous peak

What counts as a good maximum drawdown?

A lower drawdown is preferable only when the comparison is fair. A strategy with 8% drawdown and 4% return is not automatically superior to one with 12% drawdown and 30% return. The test periods, leverage, exposure and markets must also be comparable.

For an initial screening, drawdown below 10% can be viewed as relatively controlled, 10% to 20% as meaningful but potentially manageable, 20% to 30% as aggressive and above 30% as severe. These ranges are not universal professional standards. They are practical labels for discussing capital stress.

The acceptable level ultimately depends on how much loss the user can tolerate, how much uncertainty exists in the test and what action the system takes when losses exceed expectations. A drawdown that causes the operator to stop at the worst moment is not operationally acceptable, even if the strategy later recovers.

Why recovery becomes harder as drawdown grows

Losses and recoveries are not symmetrical. After a loss, the percentage gain is calculated on a smaller capital base. This is why the required recovery rises faster than the drawdown.

After a 10% decline, the account needs an 11.11% gain to return to its peak. After a 30% decline, it needs 42.86%. A 50% decline requires the remaining capital to double.

This nonlinear recovery burden is one reason drawdown control matters independently of final return. Avoiding a deep loss can preserve both capital and the time needed to rebuild it.

Intrabar drawdown compared with closed trade drawdown

A closed trade report observes results when positions are exited. It can miss a substantial temporary loss that occurred while a position remained open. Intrabar drawdown examines the adverse equity movement inside each bar and open trade.

TradingView documents an intrabar maximum drawdown calculation that considers the equity before a trade and the adverse price reached while that trade is open. The result can be higher than a simple calculation based only on completed trades.

The distinction is especially important for leveraged positions, averaging strategies and systems that hold through volatility. A strategy can close many baskets in profit while experiencing dangerous open losses along the way.

Depth is only one dimension of drawdown

Drawdown depth shows how far equity fell. Drawdown duration shows how long the account remained below its previous peak. Frequency shows how often meaningful declines occurred.

Two systems can both record a 15% maximum drawdown and create very different experiences. One may recover within ten days. The other may remain below its peak for a year. The second system ties up capital and tests confidence for much longer.

A complete report should therefore include the deepest decline, the longest recovery period, current drawdown and the distribution of other significant drawdowns rather than presenting one number in isolation.

How to compare return with drawdown

Return divided by maximum drawdown can provide a simple efficiency comparison when periods and calculation methods match. A strategy that earns 30% with 10% drawdown has a raw return to drawdown ratio of 3. A strategy that earns 40% with 30% drawdown has a ratio of about 1.33.

This ratio does not describe the order of returns, duration of losses or statistical confidence. It can also become unstable when a short test happens to contain very little drawdown. It should support analysis rather than replace it.

Profit factor, expectancy, trade count, exposure and recovery time add necessary context. A high return to drawdown ratio built from twelve trades is weaker evidence than a less spectacular result observed through many varied market conditions.

How leverage and position sizing change drawdown

If the same sequence of trades is scaled upward, both gains and losses generally become larger relative to account equity. Leverage can therefore make an acceptable strategy path intolerable without changing the entry signal at all.

Position sizing based on current equity can reduce absolute exposure after losses, while fixed notional sizing behaves differently. Pyramiding and averaging can create nonlinear exposure because the position becomes largest while price is moving against the initial entry.

A drawdown figure should never be separated from leverage, maximum position size and margin rules. The same strategy logic can show very different risk depending on how much capital each order controls.

Why the next drawdown can be larger than the backtest

A historical maximum is simply the largest decline contained in one observed sample. It is not a ceiling. Future volatility, correlation, liquidity and strategy behavior can produce a larger loss.

Backtest drawdown can also be understated by optimistic fills, missing fees, incomplete price paths, future data leakage or a test window that excludes hostile market regimes. Parameter selection can favor settings that happened to avoid past losses.

Risk planning should therefore include a buffer beyond the observed result and predefined action levels. The exact buffer cannot be derived from one maximum drawdown number alone. It should reflect test quality, sample length, strategy concentration and the consequences of model failure.

How a live system should respond to drawdown

A live system should measure equity drawdown continuously and distinguish strategy loss from deposits, withdrawals and data errors. Alerts can be triggered before a hard limit is reached.

Responses can include reducing position size, preventing new exposure, closing defined positions or pausing the strategy for review. The action should be designed before the loss occurs. Changing the rule emotionally during a decline defeats the purpose of automation.

A pause is not enough by itself. Restart conditions should define what evidence is required, who approves the restart and whether the strategy returns at full or reduced size. Drawdown control needs both a stop process and a recovery process.

How TensorParadigm evaluates drawdown

TensorParadigm evaluates return together with maximum drawdown, trade count, profit factor, exposure and the test period. A return figure without its loss path does not provide enough information to judge a system.

The evaluation also separates a historical observation from a live risk limit. Historical drawdown describes what occurred in the measured sample. A risk limit defines what the system is allowed to do now. Those numbers can inform each other, but they are not the same.

The practical objective is not to eliminate every decline. Any active trading strategy can lose. The objective is to keep losses measurable, controlled and proportionate to the evidence supporting the strategy.

Open positions are drawdown too: our live record

Drawdown should include positions that are still open. In our agents’ live record at 26 October 2026, all 78 closed trades had been profitable, so a curve built only from closed trades would show no drawdown at all.

Yet 10 trades were still open, and all were below their entry prices, by 0.7% to 11.1% before leverage. At 3× leverage, an 11% price fall would cost about a third of the margin behind that position. Those unrealized losses are part of the real drawdown.

That is why our live case study shows every open trade with its snapshot price, and why any drawdown figure should say whether open positions are marked to market.

The honest conclusion

A good maximum drawdown is not one universal percentage. It is a loss level that remains survivable, proportionate to net return and supported by a sufficiently realistic test.

Lower is generally better when everything else is equal, but everything else is rarely equal. Compare the calculation method, leverage, exposure, duration, recovery and evidence quality before comparing the headline percentages.