The Sharpe Ratio: Risk-Adjusted Return, With Math
What the Sharpe ratio actually measures, computed by hand on two twelve-month return series - plus the annualization rule, the failure modes, and how to compare ratios without being fooled.
Two funds both returned 11% last year. One did it on a smooth ride with monthly moves of one or two percent; the other swung five percent up and down on the way to the same destination. Anyone who has held both kinds of investment knows they are not equally easy to live with - but how do you put a number on that difference? The Sharpe ratio has been the standard answer since 1966: it measures how much return you earned for each unit of risk you endured to get it.
This article builds the ratio from scratch, computes it twice by hand on two twelve-month return series, and then does the more valuable work: showing what the number hides, where it misleads, and how professionals use (and misuse) it. The arithmetic is fully reproducible - every input is listed, and you can recompute the whole thing on a phone calculator.
The Definition
The Sharpe ratio is the excess return of an investment divided by the volatility of that return:
Sharpe = (Rp − Rf) ÷ σ
Three inputs, each with a definition that matters:
- Rp - the return of the portfolio or strategy over the measurement period.
- Rf - the risk-free rate: what an investment with (theoretically) no risk paid over the same period. In practice this is a short-term government T-bill yield - if T-bills pay 4%, a fund returning 4% earned nothing for the risk you took, so its Sharpe is zero.
- σ (sigma) - the standard deviation of returns: the typical size of the month-to-month (or day-to-day) swings around the average. A higher sigma means a wilder ride.
The unit is "excess return per unit of volatility". A Sharpe of 1.0 means the investment earned one percentage point of excess return for every one percentage point of standard deviation. Across long horizons, a diversified stock portfolio has historically landed somewhere in the 0.3 to 0.6 range; sustained claims above 1.0 deserve skepticism, and anything claiming 2 or 3 over long periods usually means leverage, an unusual strategy, or a measurement error.
Worked Example: Two Portfolios, Twelve Months
Assumptions: the risk-free rate is 4.0% annualized, treated as 0.333% per month (4.0 ÷ 12 - a simplification professionals refine with compounding, which changes the third decimal). Two portfolios produced these monthly returns over the past year, in percent:
| Month | Portfolio A | Portfolio B |
|---|---|---|
| 1 | +2.5 | +4.6 |
| 2 | −3.6 | −6.5 |
| 3 | +4.3 | +7.8 |
| 4 | +0.8 | +2.2 |
| 5 | −2.4 | −5.1 |
| 6 | +5.2 | +9.4 |
| 7 | −1.5 | −2.6 |
| 8 | +1.6 | +3.4 |
| 9 | +5.6 | +2.1 |
| 10 | −4.3 | −8.3 |
| 11 | +2.9 | +5.8 |
| 12 | −0.3 | +0.4 |
Step 1 - average monthly return. Portfolio A: the twelve returns sum to 10.8, so the mean is 10.8 ÷ 12 = 0.90% per month. Portfolio B sums to 13.2, mean 1.10% per month. Annualized, A earned about 10.8% and B about 13.2% - B looks like the better fund.
Step 2 - standard deviation. Subtract the mean from each month, square the differences, average the squares, and take the square root. For A this gives σ = 3.37% per month; for B, σ = 5.68% per month. (Spreadsheet check: the STDEV.S function on the two columns returns exactly these values.)
Step 3 - excess return. Subtract the monthly risk-free rate: A: 0.90 − 0.333 = 0.567%. B: 1.10 − 0.333 = 0.767%.
Step 4 - divide and annualize. Monthly Sharpe for A: 0.567 ÷ 3.37 = 0.168. For B: 0.767 ÷ 5.68 = 0.135. Annualize by multiplying by the square root of 12 (≈ 3.464):
| Measure | Portfolio A | Portfolio B |
|---|---|---|
| Average monthly return | 0.90% | 1.10% |
| Annualized return | 10.8% | 13.2% |
| Monthly standard deviation | 3.37% | 5.68% |
| Annualized volatility | 11.7% | 19.7% |
| Monthly Sharpe | 0.168 | 0.135 |
| Annualized Sharpe | 0.58 | 0.47 |
The answer inverts the naive ranking. B earned 2.4 percentage points more per year - and delivered less return per unit of risk. Whether that trade is worth it depends entirely on the investor, which is the point: the Sharpe ratio does not rank investments as good or bad, it standardizes them so you can price the ride.
The √12 rule and where it comes from. Monthly Sharpe ratios are annualized by multiplying by √12, not 12, because independent monthly volatilities add in variance, not in standard deviation - and variance grows twelvefold per year while standard deviation grows by its square root, ≈3.46. Return averages do scale linearly; volatility does not. Using 12 instead of √12 is one of the most common errors in amateur backtests, and it inflates annualized Sharpe by a factor of about 3.5.
What the Sharpe Ratio Hides
- It treats upside and downside identically. Sigma measures deviation from the mean in both directions, so a fund with huge pleasant surprises scores worse than a steady one. That rewards smoothness even when the smoothness comes from strategies that quietly sell insurance against rare disasters - the notorious pattern of strategies with beautiful Sharpe ratios for years, followed by one catastrophic month. The Sortino ratio, which uses only downside deviation, is the common refinement.
- It assumes returns are roughly normal. Market returns have "fat tails" - crashes happen more often than a bell curve predicts. A single-year Sharpe computed on calm data systematically understates tail risk.
- It is a period statistic. A Sharpe computed over one bullish year tells you almost nothing; over ten years including a crash, it tells you something real. Always ask how many months - and which months - went into the number.
- Leverage can manufacture it. Borrowing to hold a larger, safer position mechanically raises Sharpe without raising skill. A levered bond portfolio can show a higher Sharpe than unlevered stocks. Higher Sharpe means more return per unit of volatility, not more wealth.
A Leveraged Twist: Why Borrowing Does Not Raise the Sharpe
Here is the calculation that explains why "higher Sharpe" and "more money" are different things. Take Portfolio A and run it at 1.5 times size: borrow an extra half position at the risk-free rate. The monthly mean becomes 1.5 × 0.90% = 1.35%, minus the borrowing cost of 0.5 × 0.333% = 0.167%, for a net mean of 1.183%. The volatility scales too: 1.5 × 3.37% = 5.05%.
Now compute the Sharpe: (1.183 − 0.333) ÷ 5.05 = 0.168 monthly - exactly what unleveraged A earned. Leverage at the risk-free rate stretches both axes of the ratio by the same amount, so the ratio itself is unchanged; what changes is the absolute return and the absolute pain. This is the logic professionals mean by "Sharpe is scale-invariant", and it is why a fund cannot raise its Sharpe simply by gearing up. (The claim only breaks when borrowing costs exceed the risk-free rate or leverage introduces new risks - both of which happen in practice.)
Getting the Inputs From Real Data
Everything above uses a spreadsheet's worth of monthly returns, and that is genuinely all it takes to reproduce the number for any fund you own. The monthly returns are published in every fund fact sheet and in free charting tools (most brokerages display a monthly total-return history on request); the risk-free rate is the published yield on short-term Treasury bills, available from the Treasury's own website; and the standard deviation is one function call - STDEV.S in Excel or Sheets - applied to the same monthly column. Ten minutes, no subscription required.
Three hygiene rules keep the result honest: use total returns (with dividends reinvested), not price returns, or you will understate index funds and overstate volatility-adjusted differences; use the same frequency for return and risk-free inputs; and count the months - a Sharpe from 12 observations is a sketch, one from 60 is a measurement, and the difference should temper how much any single number moves a decision.
How to Use It Without Being Fooled
The ratio earns its keep in three comparisons. First, across time for the same strategy: if a fund's five-year Sharpe was 0.55 and the last two years show 1.4, ask what changed - usually either a favorable market regime or a change in what the fund actually holds. Second, between strategies at the same risk level: at similar sigma, the higher Sharpe genuinely earned more. Third, between asset classes: stocks and bonds have very different typical Sharpe ranges, which is exactly the arithmetic behind blending them - a mix can have a higher Sharpe than either component alone because the two volatilities partially cancel, the same mechanism explained in the asset allocation guide.
What not to do: chase the highest Sharpe number in a screener. Funds optimized for the ratio cluster in smooth-looking strategies that carry hidden tail risk, and the metric says nothing about fees, taxes, or whether the strategy still works. Treat it as one lens - the volatility-adjusted one - and pair it with maximum drawdown and the tracking-difference analysis for index funds.
A Caveat on the Risk-Free Rate
The "risk-free" input is a choice, and it moves the number. T-bill yields, money market rates, and deposit rates can differ by fractions of a percent - small next to market volatility, but enough to shift a reported Sharpe by 0.02-0.05. When comparing two published Sharpe ratios, confirm they used the same measurement period, return frequency, and risk-free rate; ratios computed by different data providers on the same fund routinely differ by more than the differences people build decisions on.
This article is educational arithmetic, not investment advice: it describes a metric, not a recommendation to buy, sell, or hold anything, and past Sharpe ratios have never guaranteed future ones. For the neighboring concepts, the risk-and-return basics in the asset allocation guide and the volatility mechanics in the bond price guide cover how σ behaves across asset classes, and the behavior gap article explains why the investor's own Sharpe - returns actually experienced - is usually lower than the fund's.
