Module 1 of 14 · Foundations

Measuring a strategy — more than one number

Before building anything, decide how you will judge it. A systematic trader backtests dozens of ideas, and the yardstick you pick quietly decides which ones survive. Use one that rewards the wrong thing and you will select the wrong strategy with great confidence.

The workhorse is the Sharpe ratio: average annual return divided by annual standard deviation of returns. It answers “how much return per unit of risk?” and, because futures need very little cash, it is the number that matters — you can dial risk up or down with leverage, so the return per unit of risk is what you are really buying. Annualise daily numbers by multiplying the mean by about 256 business days and the standard deviation by √256 = 16.

But standard deviation treats a sudden 10% loss and a sudden 10% gain as equally “risky”, and it says nothing about how the returns arrive. Three more measures fill the gaps:

  • Skew — is the distribution lopsided? Negative skew means steady small gains and occasional large losses (selling insurance). Positive skew is the reverse (buying it). Measure it on monthly returns; daily skew is dominated by a handful of freak days.
  • Tail ratios — compare an extreme percentile (1st) with a moderate one (30th) and divide by what a normal distribution would give. A value of 2 means your bad days are twice as bad as a bell curve predicts.
  • Drawdown — the fall from the previous peak. The average drawdown is a steadier statistic than the single worst one, which depends heavily on how long your history is.

🎛 Return-shape lab

0.7
-0.6

Sharpe (both)

0.70 / 0.70

plain / shaped

Monthly skew

-0.11 / -1.11

negative skew

Worst drawdown

41% / 30%

of capital

Worst single day

-3.5 / -12.9

in daily std devs

Both lines are built to have the same annual volatility (20%) and the same Sharpe ratio over the full 15 years. The only difference is the shape of the daily returns: the grey line is bell-shaped, the amber line adds rare jumps in one direction and is rescaled to the same mean and volatility. Skew is measured on monthly returns; the worst day is shown in units of a normal day's standard deviation (a bell curve almost never exceeds 4–5). Synthetic data.

Same Sharpe, different strategy

Drag the jump slider to −1. The two equity curves have the same Sharpe ratio and the same volatility, yet the negatively skewed one climbs smoothly and then falls off cliffs: monthly skew turns sharply negative and its worst day is a loss that a bell curve says should essentially never happen. That is the profile of strategies that sell insurance: carry trades, short volatility, and mean reversion. Notice also that the worst drawdown is not a reliable guide here — the plain line can show the deepest drawdown purely by the luck of its path, which is why average drawdown and skew are steadier measures.

Positive skew — trend following is the classic example — loses slowly and wins suddenly. It is psychologically harder to stick with in the dull stretches but much safer when markets break. A fair comparison of two strategies therefore reports Sharpe and skew and tails, and a sensible portfolio blends both shapes rather than maximising Sharpe alone.

Things to try

  • • Set Sharpe to 0 and look at the curves: a zero-edge strategy still produces long, convincing-looking runs. Twenty years is not as much evidence as it feels like.
  • • Push the jump slider to +1: skew flips positive and the worst day becomes an ordinary one, even though volatility is identical.
  • • Find the jump size where the worst day first exceeds 10 standard deviations. A risk model built on volatility alone would call that day impossible — which is why rare-event risk needs its own controls (Module 14).

Original interactive lesson. Labs run on a seeded synthetic market in your browser, built to illustrate mechanics, not to predict any real market. Contract figures are illustrative; check exchange specifications. Educational analysis, not investment advice.

How are we doing?