Every trading platform, strategy marketplace and fund factsheet reports performance metrics. They are useful summaries, but each compresses a complex history into a single number, and each can be flattered by choices about sample, frequency and presentation.
Total and annualised return
Returns are the headline, but the period matters enormously. A strategy that started at a market low will look better than the same strategy started at a high. Always ask for the full history and the start date.
Volatility and the Sharpe ratio
The Sharpe ratio divides excess return by volatility, rewarding smooth returns. It assumes volatility captures risk, which fails for strategies with rare but large losses — such as selling options — that can show a high Sharpe ratio until the day they do not. Short samples and monthly smoothing also inflate it.
Maximum drawdown
Maximum drawdown measures the largest peak-to-trough fall. It is closer to how investors experience risk and harder to game than volatility. Pair it with the time taken to recover: a 20% drawdown recovered in two months is very different from one that lasts two years.
A high Sharpe ratio over a short history is a hypothesis, not a result.
Other useful measures
- Sortino ratio — like Sharpe, but penalises only downside volatility.
- Calmar ratio — annual return divided by maximum drawdown.
- Win rate and payoff ratio — meaningful only together; a high win rate with small wins and large losses can still lose money.
- Exposure and turnover — essential context for judging costs and capacity.
Questions to ask of any track record
- Is it live, simulated or backtested — or a blend?
- Are costs, fees and slippage included?
- How many strategies were tested before this one was shown?
- Does the history include at least one difficult market period?
The multiple-testing problem
The single most important question about any performance record is how many alternatives were tried before it was selected. If a researcher tests a hundred strategy variants on the same data, the best one will look impressive by chance alone. Academic work on the 'deflated' Sharpe ratio formalises this intuition: the more trials, the higher the bar a result must clear to be considered meaningful.
Published track records almost never disclose how many trials preceded them. Readers should assume the number is larger than one and discount accordingly.
Live versus simulated records
A live record, even a short one, carries information a backtest cannot: real fills, real costs, and the discipline of decisions made without hindsight. A blended record that splices a backtest onto a short live period should be read with the boundary clearly in mind — and the live portion weighted far more heavily.
Comparing like with like
Metrics are only comparable when computed the same way. Daily versus monthly returns, inclusion or exclusion of fees, and the choice of risk-free rate all change the numbers. When comparing strategies or products, recompute metrics from the underlying return series where possible rather than relying on headline figures.