Drawdown

import Figure from ’../../components/charts/Figure.astro’; import LineChart from ’../../components/charts/LineChart.astro’; import { drawdownRecoveryLine } from ’../../data/sample-data’;

Drawdown is the peak-to-trough decline in an account’s equity, expressed as a percentage of the peak. If a $10,000 account drops to $7,500 from a previous high, the drawdown is 25%. Drawdown is the single most important risk metric for any trading account, because the math of recovering from it is non-linear and unfriendly.

The asymmetry

The function that connects “loss” to “gain needed to recover” is y = 1 / (1 - x) - 1. It is convex, which means it goes non-linear fast.

• 10% drawdown requires 11.1% gain to recover.
• 20% drawdown requires 25%.
• 30% drawdown requires 42.9%.
• 50% drawdown requires 100%.
• 80% drawdown requires 400%.
• 90% drawdown requires 900%.

The reason matters. A 50% loss leaves only half the original capital. To get back to the original amount, the remaining half has to double. Each subsequent percentage point of loss disproportionately increases the gain required from the smaller remaining base. The curve is structurally biased against deep drawdowns.

Why this drives sizing

If your maximum acceptable drawdown is around 25%, the position-sizing math has to support that limit. Risking 5% per trade means five consecutive losses (well within the binomial range for any normal-win-rate strategy) takes you to roughly the limit in one sitting. The 1% per-trade convention exists precisely to keep the worst-case streak within recoverable range.

The fixed-percent rule (typically 0.5-2% of equity per trade) has this property baked in: as the account draws down, the absolute dollar risk shrinks automatically. The strategy de-leverages into losses, which means the equity curve cannot slide off the convex part of the recovery curve in a single bad streak.

Two flavours of drawdown

The single number “30% drawdown” can hide two different distributions:

• Drawdown depth. The worst peak-to-trough percentage hit in the sample. The number above.
• Drawdown duration. How long the account spends below its previous peak before recovering. A 20% drawdown that recovers in two months is qualitatively different from a 20% drawdown that takes four years to recover (which is the typical pattern for long-trend-following CTAs after a regime shift).

A trader running a trend strategy should expect both kinds. A trader running a high-frequency strategy that genuinely works should expect shallower depth but more frequent occurrence. The two patterns demand different psychological tolerances.

What it does and does not tell you

A backtest reporting “maximum drawdown 18%” is reporting a sample statistic, not a worst case. The next regime might produce 25%, or 35%, or 50%. The drawdown the strategy realises after you start trading it is almost always larger than the worst drawdown in the historical backtest, because the backtest sampled a finite history and the future continues to deliver new realisations.

The honest position is to size positions assuming the realised drawdown will be roughly 1.5x to 2x the backtested worst case. That margin exists for the regimes the backtest did not include.

See Risk Management Basics for the position-sizing framework drawdown drives, and Currency Correlation and Hidden Risk for why portfolio-level drawdowns are usually larger than the per-trade arithmetic would suggest.