The moment a backtest looks convincing until max drawdown enters the room: a beginner's guide to testing survival, not just returns

A backtest becomes useful only when it shows whether the strategy could survive its losses. Total return may attract attention, but max drawdown reveals the deepest historical fall from an equity peak to the next trough.

Read the equity curve before the return figure

Suppose a backtest turns $10,000 into $16,000 over three years. A 60% gain sounds convincing until the equity curve shows that the account reached $13,000, fell to $7,800, and later recovered.

That decline is a 40% max drawdown:

`($13,000 − $7,800) ÷ $13,000 = 40%`

Could you continue following the strategy after watching $5,200 disappear from its peak value? Could your account still meet margin requirements? Would one more losing trade force you to reduce positions or stop?

Those questions test survival. The final return figure cannot answer them.

Calculate drawdown from every new peak

Drawdown measures the percentage decline from the highest account value recorded so far. Each time the equity curve reaches a new peak, that value becomes the reference point.

For a simple backtest:

1. Record the account value after every trade, including fees.
2. Track the highest value reached up to each point.
3. Calculate the percentage decline from that peak.
4. Find the largest decline in the full test.

Use account equity rather than the price of the traded asset. A strategy can hold an asset through a 20% decline while the account falls only 4% because the position was small. Position size determines how market movement reaches the portfolio.

If your backtesting tool reports max drawdown automatically, verify whether it uses closed trades only or marks open positions to market. Closed-trade calculations can conceal losses that existed while a position remained open.

Translate drawdown into recovery required

Losses and recoveries are asymmetric. A 10% loss requires an 11.1% gain to return to the starting value. A 25% loss requires 33.3%. A 50% loss requires 100%.

That difference changes how you should compare strategies. Consider two hypothetical results:

• Strategy A returns 18% per year with a 12% max drawdown.
• Strategy B returns 24% per year with a 45% max drawdown.

Strategy B has the higher historical return. It also requires a much larger recovery after its worst decline, and its position sizes may become harder to maintain as capital contracts.

Do not select between them from those two figures alone. Examine the length of each drawdown, the number of losses inside it, and whether the strategy depended on a few unusually profitable trades.

Test conditions that the clean backtest omitted

A backtest can look precise while relying on generous assumptions. Add the costs and constraints you would face in actual trading:

• Apply commissions, spreads, slippage, and funding costs where relevant.
• Delay entries to reflect when the signal would have been available.
• Reject trades that exceed realistic liquidity or buying power.
• Model gaps beyond the planned stop price.
• Limit position size using the capital available at that point in the test.
• Include open-position losses when calculating account equity.

Position sizing deserves particular attention. If a stop widens from $0.20 to $0.50 while the share count stays fixed, the amount at risk rises by 150%. The safer response may be a smaller order, as explained in what happens when the opening candle widens your risk.

These adjustments usually make the result less attractive. That is useful. A weaker result based on plausible execution conditions offers more information than a polished result you could never reproduce.

Separate development data from evaluation data

Testing dozens of ideas on one dataset encourages overfitting. Eventually, one combination will look impressive by chance.

Divide the available history into separate periods. Use the first period to develop the rules, then freeze those rules and run them on data that did not influence the design. This second period is the out-of-sample test.

Also divide results by market condition. A strategy that survives a steady rise may fail during a sharp decline, a low-volatility range, or an overnight gap. Check different assets and timeframes only when the strategy was designed to trade them. Expanding the test after seeing weak results can become another form of curve fitting.

Stress the order of wins and losses

Historical trades arrived in one sequence. A different sequence could produce a deeper drawdown even if the individual trade outcomes stayed the same.

One basic stress test is to reshuffle the order of completed trades many times and calculate max drawdown for each sequence. This does not predict the future, but it exposes sequence risk. If the original backtest shows a 14% max drawdown while reordered sequences regularly reach 25%, planning around 14% would leave little margin for error.

You can also weaken individual assumptions. Increase slippage, remove the five best trades, reduce profit targets slightly, or place several losses together. A strategy that collapses after a small change is fragile.

Set a drawdown rule before trading

Choose a drawdown threshold based on capital constraints and personal tolerance before money is at risk. Then define what happens as the account approaches it.

For example, an illustrative plan might reduce position risk after an 8% drawdown, pause new entries at 12%, and require a full review before trading resumes. The exact figures must reflect the account, instruments, and tested strategy. They are controls, not universal recommendations.

Record the rule in a trading journal. Include the backtest assumptions, expected drawdown range, position-sizing method, and conditions that invalidate the strategy. If an AI system generates a trade idea, treat it as a proposal to review against those limits. Approval should depend on current portfolio risk, not the historical return printed at the top of a report.

Your next step is concrete: reopen one backtest, add realistic costs, calculate peak-to-trough drawdown using marked-to-market equity, and remove the five best trades. If the strategy still fits your predetermined loss limit, it has earned further testing with paper trades or minimal size.

Educational content, not financial advice.