This article explains a risk-adjusted return metric and the ways it gets gamed. The worked numbers are arithmetic illustrations, not performance claims. Nothing here is investment advice. Decisions are yours.
The Sharpe ratio is return per unit of risk: a strategy's excess return divided by the volatility of that return. It is the industry's standard one-number summary, and in crypto it arrives with a built-in flattery few sellers disclose: a 365-day market annualizes the same daily behavior to a number roughly 20% higher than the equity convention would print.
The definition and what it buys
William Sharpe proposed the measure in 1966. Take a return stream, subtract the risk-free rate, divide by the standard deviation of the returns. The division is the entire point: raw return can always be manufactured by adding risk, so a doubling bought with triple the volatility is a worse deal per unit of exposure than it looks. Sharpe puts every strategy on a common risk-adjusted scale, which is why allocators, and this site's own gates, speak in it rather than in raw multiples. A return figure without a risk denominator is marketing, the same trick behind the 95% lie.
The 365-day flattery
Sharpe ratios are quoted annualized, and annualization multiplies the per-period ratio by the square root of the number of periods in a year. Equities have roughly 252 trading days; crypto trades every day of the calendar. The square root of 365 over 252 is about 1.20, so identical daily risk-adjusted behavior prints an annualized Sharpe about 20% higher under crypto's convention than under the equity convention, one more consequence of a 365-day market. Nothing dishonest has happened in the arithmetic; the dishonesty arrives when a crypto product quotes its Sharpe next to equity benchmarks without mentioning the convention gap, harvesting a fifth of a Sharpe point from the calendar. Any cross-market comparison should state its day count, and a seller who will not is answering a different question, so put them through seven concrete checks.
What 1, 2, or 3 actually means
The folklore bands: around 1 is respectable for a liquid directional strategy, around 2 is strong, 3 and above is exceptional, and, at any meaningful scale and history, suspicious. The bands earn their keep through the statistics underneath: an annualized Sharpe times the square root of the track's length in years approximates a t-statistic, the significance of the whole record. A Sharpe of 1.0 sustained for two years is a t-stat near 1.4, genuinely weak evidence, indistinguishable from luck at conventional thresholds. The same Sharpe over four years crosses 2 and starts meaning something. Crypto's short usable histories make this the binding constraint: most strategies here cannot yet have statistically meaningful Sharpes at all, and a two-year backtest with a Sharpe of 3 is a smaller miracle than it appears once you count how many two-year windows and parameter sets were searched to find it, the trial-count arithmetic at the heart of overfitting in crypto.
Where Sharpe misleads in this market
Four crypto-specific failure modes. Fat tails: standard deviation understates the risk of a distribution whose worst days are liquidation-cascade days, so two strategies with equal Sharpes can carry very different ruin profiles. Upside penalty: Sharpe's denominator counts violent gains as risk, so a momentum system whose best months are explosive gets marked down for the very asymmetry it exists to capture; the downside-only alternative is the subject of the Sortino piece. Survivorship inflation: a backtest computed on today's coin list skips the crypto graveyard, inheriting phantom return with no matching phantom volatility, and its Sharpe inflates accordingly. And regime concentration: a Sharpe earned entirely inside one bull phase is a description of the phase, not the strategy, because everyone is a genius in a bull market.
How our own audits treat it
The house position: Sharpe is necessary and insufficient. Our own audit ran the equity research's headline result through the deflated Sharpe ratio, a correction that discounts a Sharpe for the number of strategy variants tried, the track length, and the return distribution's skew and tails, and through a formal probability-of-overfitting test across the full trial archive. The crypto derivation inherits those gates pre-registered: no branded score ships without clearing them in public, run by run, in the derivation journal. Until then, this site publishes measurements rather than a performance record, so there is no Sharpe to sell you, and that is the point: when one appears, it will arrive with its day-count convention, its trial count, and its deflation stated, because a Sharpe without those three footnotes is a number wearing a costume. Decisions are yours.