0bps

Learn / Track III / lesson 5 of 5

Drawdowns and position sizing

The risk number traders actually feel — how deep losses compound, how to size so you survive them, and how long recovery takes.

9 min read

Definition

Track the running peak of equity, Mt=max⁡s≤tEsM_t = \max_{s \le t} E_s. The drawdown at time tt is

Dt=Mt−EtMtD_t = \frac{M_t - E_t}{M_t}

and the maximum drawdown is the largest DtD_t over the sample. The drawdown duration is how long equity spends below a prior peak.

The asymmetry of losses

After losing a fraction dd, you need a gain of

11−d−1\frac{1}{1 - d} - 1

to get back to even. A 10% loss needs 11%; a 25% loss needs 33%; a 50% loss needs 100%; a 90% loss needs 900%. Deep drawdowns are not just painful, they are mathematically expensive. Explore it in the drawdown calculator.

How long does recovery take?

If a strategy compounds at gg per year, recovering from drawdown dd takes about

T≈−ln⁡(1−d)gT \approx \frac{-\ln(1 - d)}{g}

years. At g=15%g = 15\%, a 30% drawdown costs roughly 2.4 years of compounding just to stand still.

Expected maximum drawdown

For a strategy with annualised Sharpe SRSR and volatility σ\sigma, max drawdown grows with the length of the sample, and is larger the lower the Sharpe. Simulating ten years of a 20%-vol strategy, the median maximum drawdown is about 1.5σ1.5\sigma (30%) at Sharpe 1 and about 2σ2\sigma (40%) at Sharpe 0.5 — and one path in four does noticeably worse. A 40% drawdown in a Sharpe-0.5 strategy is the expected outcome, not proof the strategy is broken.

Sizing per trade

The most common practical rule: risk a fixed fraction of equity on each trade, measured from entry to stop.

position size=equity×risk fraction∣entry−stop∣\text{position size} = \frac{\text{equity} \times \text{risk fraction}}{|\text{entry} - \text{stop}|}

With 1% risk per trade, ten consecutive losers cost about 9.6% (compounded), not 10%, and you are still in the game. With 10% risk, the same streak costs 65%.

Worked example

Equity 20,000, risk 1% (200), long BTC at 64,000 with a stop at 62,400. Distance is 1,600, so size is 200/1600=0.125200/1600 = 0.125 BTC, a notional of 8,000 (0.4× leverage). Use the position sizer to do this per trade.

Volatility-scaled sizing

Using ATR (average true range) as the stop distance — a stop at, say, 2×ATR2 \times \text{ATR} — makes position size inversely proportional to volatility. Each trade then carries a similar amount of risk regardless of the regime. This is what many vault specs mean by “ATR position sizing”.

Common mistake

Judging a strategy by its backtest’s maximum drawdown and expecting the live drawdown to stay inside it. The backtest shows one path. Over a longer live period, a deeper drawdown is close to certain.

Put it to work

esc

Try , , or .

↑↓ move↵ open