Two traders finish the year up the same amount. One got there on a smooth ramp; the other on a heart-attack rollercoaster. The Sharpe ratio is the number that finally tells them apart.
What Sharpe measures: excess return per unit of volatility
The Sharpe ratio answers one question: how much return did you earn for each unit of risk you took, where risk is measured as volatility — the standard deviation of your returns.
The formula is:
Sharpe = (mean return − risk-free rate) / standard deviation of returns
The numerator is your excess return — what you made above a do-nothing benchmark like T-bills. The denominator is how bumpy the ride was. A higher Sharpe means you extracted more reward per unit of turbulence. It rewards consistency and punishes wild swings, which is exactly the trait a prop firm cares about, because a smooth equity curve is a curve that does not breach a trailing drawdown by accident.
Computing Sharpe from daily or per-trade returns
You can build Sharpe from either daily account returns or per-trade returns. Daily is usually cleaner because it captures the real experience of holding the account through time, including days you did nothing.
The steps for a daily series:
- Compute each day’s return as a percentage of account equity.
- Subtract the per-period risk-free rate (often set to zero for short intraday horizons — just be consistent).
- Take the mean of those excess returns.
- Take the standard deviation of the same series.
- Divide mean by standard deviation. That is your per-period Sharpe.
Per-trade Sharpe uses each trade’s return in R or percent instead of each day. It is useful for comparing setups but it hides idle time and clustering, so treat it as a strategy-level diagnostic, not an account-level score.
Annualizing correctly and the sqrt-of-N trap
A daily Sharpe is small and hard to compare. To annualize, you multiply by the square root of the number of periods per year:
Annualized Sharpe = per-period Sharpe × √N
For daily data, N is roughly the number of trading days in a year, so you multiply by about √252 ≈ 15.9. For weekly data you use √52.
The sqrt-of-N trap catches everyone at least once. Two mistakes dominate:
- Mismatched N. Annualizing a weekly Sharpe with √252 inflates the number absurdly. Match N to the sampling frequency of your returns.
- Autocorrelation. The √N rule assumes each period is independent. If your returns trend or mean-revert day to day — common for momentum and grid strategies — annualizing overstates or understates the true figure. Take a big annualized Sharpe from a short, correlated sample with heavy skepticism.
A Sharpe computed on twenty trades is barely a rumor. This is the same small-sample problem Shibiki addresses by scoring live edge health with a Wilson confidence interval — the band stays wide until you have enough data for any ratio to mean something, so you are not fooled by a two-week hot streak dressed up as a high Sharpe.
Why Sharpe punishes upside volatility unfairly
Here is Sharpe’s built-in flaw: standard deviation treats a big winning day exactly like a big losing day. Both increase volatility, both lower your Sharpe. A trader who occasionally catches a monster 6R runner is penalized for it, because that upside spike widens the denominator.
That is nonsense from a trader’s point of view. Nobody has ever been hurt by an unusually large profit. Sharpe’s symmetry means a strategy with rare, huge winners can score worse than a dull one that never surprises you — even though the first is more profitable and, in the ways that matter, safer. This single blind spot is the reason the next two ratios exist.
Benchmarks: what Sharpe separates good from great
Sharpe is best read as a relative and directional number, not an absolute grade. Rough industry intuition, applied to an annualized figure on a credible sample:
- Below 1 — returns are not clearly compensating for the risk taken.
- Around 1 — respectable; a real, tradeable edge.
- 2 and up — strong, if it survives out-of-sample.
- 3+ — exceptional, and worth double-checking for the sqrt-of-N trap, autocorrelation, or an accidentally short sample before you believe it.
Compare Sharpe against your own history first. A rising Sharpe across quarters is the signal; a single quarter’s number in isolation is mostly noise.
When to prefer Sortino or Calmar instead
Because Sharpe mispriced your upside, two alternatives fix different pieces of it:
| Ratio | Denominator (risk measure) | What it penalizes |
|---|---|---|
| Sharpe | Standard deviation of all returns | All volatility, up and down |
| Sortino | Downside deviation only | Only returns below your target |
| Calmar | Maximum drawdown | Worst peak-to-trough loss |
Use Sortino when your strategy has healthy positive skew and you are tired of being punished for big winners. Use Calmar when survival is the constraint — which, for a funded trader living under a trailing drawdown, it always is. Calmar speaks directly to the question a prop firm is really asking: how deep is the worst hole this account ever dug?
Report Sharpe alongside expectancy, not instead of it. An expectancy calculator gives you the per-trade edge; Sharpe tells you how smoothly that edge arrived. Shibiki auto-journals your fills and computes these together, so risk-adjusted return stops being a spreadsheet chore and becomes something you actually glance at before sizing the next trade.
Related: Expectancy calculator · Trailing drawdown · Shibiki vs Edgewonk