
Risk-Adjusted Returns: Sharpe Ratio, Sortino, and Alpha
Measuring how much return you get per unit of risk taken
Return Without Context Misleads
Two strategies can post the same annual percentage while exposing you to wildly different drawdowns. Risk-adjusted metrics ask a disciplined question: how much pain did you take per unit of reward? They do not replace judgment — they stop headline numbers from hiding leverage, concentration, and gap risk.
Start with a simple split: excess return versus a baseline (often a short-term government yield), and risk measured as dispersion of returns or sensitivity to a benchmark.
Sortino and Treynor: Different Denominators
The Sortino ratio uses downside deviation in the denominator — volatility of returns below a target or zero — so positive surprises are not “punished” as risk. The Treynor ratio divides excess return by beta, emphasizing compensation for systematic exposure rather than total volatility.
Neither is universally “better.” Pick the denominator that matches the risk you actually fear.
Alpha and Information Ratio
Jensen’s alpha (in a CAPM framing) measures return above what beta would predict. The information ratio scales active return by tracking error — how consistently you deviate from a benchmark. A noisy strategy with occasional jackpots may look worse than a steadier one when you care about implementation risk.
Applying Metrics in Crypto
Crypto return series are non-normal: fat tails, regime shifts, and funding mechanics on perpetuals. Report both rolling Sharpe and maximum drawdown. If you trade on GaiaEx, include fees, funding, and liquidation proximity in your simulation — otherwise your “alpha” is a spreadsheet fantasy.
A Practical Checklist
- Define the return series (log vs simple) and stick to it.
- Pick a risk-free proxy that matches your horizon.
- Document whether statistics are full-sample or rolling.
- Never optimize purely for Sharpe in-sample without a holdout regime.