Excess return per unit of total risk: (return - risk-free rate) divided by the volatility of returns, annualized. The standard score for risk-adjusted performance.
Definition
Sharpe = (R_strategie - R_risikofrei) / sigma_strategie
annualized via Sharpe_annual = Sharpe_pro_Periode × sqrt(Perioden pro Jahr).
Developed by William F. Sharpe (1966). Measures how much excess return a strategy delivers per unit of variability. Two strategies with the same return: the calmer one (lower vol) has the higher Sharpe and is the better one.
Interpretation
| Sharpe (annualized) | Assessment |
|---|---|
| < 1.0 | meager |
| 1.0 – 2.0 | good |
| 2.0 – 3.0 | very good |
| > 3.0 | excellent — overfit-suspect in backtests |
How Botty uses it
Sharpe is the central risk-adjusted metric of the scouts and the edge correlation audit (edge_correlation_audit.py): funding carry emerged there as the risk-adjusted crown jewel (Sharpe 3.32 at 2.1% vol). The indicator_lab additionally reports a per-trade Sharpe as a measure of signal quality. When bundling edges, weighting is done by Sharpe contribution.
Limitations
- Penalizes upside and downside vol equally. A strategy with explosive gains gets punished for its "volatility". ⇒ Sortino (downside only) / Calmar (drawdown) as complements.
- Assumes a normal distribution. With fat tails and skew (e.g. short-vol / carry / option selling), the Sharpe looks dreamy — until the tail arrives. A high Sharpe from collecting small premia masks crash risk.
- In-sample overestimates. The best of N tested strategies has a high Sharpe by pure chance. The antidote: the Deflated Sharpe Ratio.