
Updated November 2026
The Sharpe ratio measures how much return a strategy earned above a risk-free rate for each unit of volatility it took on. You subtract the risk-free rate from the average return and divide the result by the standard deviation of those returns, so a higher number means more excess return per unit of risk over the period measured. William F. Sharpe introduced it in 1966 under the name "reward-to-variability ratio".
Most explanations stop at the formula. This guide runs it step by step on two hypothetical strategies, then runs it again after an illustrative trading cost, and the ranking reverses. Every number comes from a short calculation script. Nothing here is a forecast, a recommendation or a record of real results.
What is the Sharpe ratio?
The Sharpe ratio answers one question: was the extra return worth the swings? Of two strategies that finish a year level, it scores the one with steadier months higher.
It has three inputs:
- The strategy's returns, period by period. This guide uses monthly returns.
- A risk-free rate, the return on a near-riskless asset over the same periods. In Sharpe's words, the original benchmark was a riskless security, so the top of the ratio is the excess return over that asset.
- The standard deviation of excess returns, which measures how widely they spread around their average. It is the usual statistical measure of volatility.
Sharpe first set out the measure in "Mutual Fund Performance" (Journal of Business, January 1966). In a 1994 article he notes that the measure caught on but his name for it did not. Today it is a standard tool for how trading strategies are compared.
The Sharpe ratio formula, with a worked example
The formula is:
Sharpe ratio = (average return - risk-free rate) / standard deviation of excess returns
Sharpe's 1994 article describes the historic version as the average of the excess returns divided by their standard deviation, and gives the spreadsheet form AVERAGE / STDEV. STDEV is the sample standard deviation, which divides by the number of observations minus one. The worked example follows that method.
Step 1: the inputs
Two hypothetical strategies, invented for this guide. Strategy A is a frequent trader that opens and closes 12 positions a month. Strategy B is a slower trend-following strategy that completes 2 round trips a month. Their returns are gross, before any trading cost.
Month | Strategy A, gross return (%) | Strategy B, gross return (%) |
|---|---|---|
1 | 1.8 | 4.2 |
2 | 0.4 | -2.1 |
3 | 2.1 | 3.5 |
4 | 1.5 | 1.0 |
5 | -0.6 | -3.4 |
6 | 1.9 | 5.1 |
7 | 0.9 | 0.6 |
8 | 2.6 | -1.2 |
9 | 0.2 | 4.4 |
10 | 1.4 | 2.3 |
11 | -0.3 | -0.8 |
12 | 2.5 | 2.0 |
Illustrative only. Hypothetical return series, not a backtest, not a track record and not Ouinex results.
Step 2: the risk-free rate
For a UK reader, the Bank of England's Bank Rate is a natural reference. The Bank raised it to 4% on 5 November 2026. Before that it had stood at 3.75% since 18 December 2025.
To keep the arithmetic visible, the example uses 4.00% a year as a constant and divides it by 12: 0.33% a month (0.3333% unrounded). A real backtest would use the rate in force in each month. Whichever rate you pick, use the same one for every strategy you compare.
Step 3: average, standard deviation and the monthly ratio
For strategy A, the 12 returns add up to 14.4%, so the average is 1.20% a month. Subtracting 0.33% gives an average excess return of 0.87%. The monthly excess returns run 1.47, 0.07, 1.77, 1.17, -0.93, 1.57, 0.57, 2.27, -0.13, 1.07, -0.63 and 2.17. Their sample standard deviation is 1.07%. Dividing gives the monthly Sharpe ratio: 0.8667 / 1.0728 = 0.8079.
For strategy B, the average is 1.30% a month and the average excess return is 0.97%. The swings are much larger, so the standard deviation is 2.76%. The monthly Sharpe ratio is 0.9667 / 2.7581 = 0.3505.
Step 4: annualise with the square root of 12
Monthly ratios are usually converted to annual ones so that results can be compared. The rule is to multiply by the square root of 12, which is 3.4641.
The reason is how the two parts of the ratio grow with time. If monthly excess returns are independent of each other, the annual average is 12 times the monthly average, but the annual standard deviation is only the square root of 12 times the monthly one. Twelve divided by the square root of 12 is the square root of 12. Sharpe's 1994 article sets out this case, which assumes zero serial correlation between periods, and notes that compounding and correlated returns make real data more complicated.
- Strategy A: 0.8079 x 3.4641 = 2.80
- Strategy B: 0.3505 x 3.4641 = 1.21
Before costs, A looks far better. A figure as high as 2.80 from a single year should prompt two checks: what the trading cost, and how long the record is.
The same two strategies after costs
Every trade costs something: spreads, commissions, slippage. A strategy that trades six times as often pays six times as much.
Assumption, for illustration only: trading costs of 0.05% of the position's value per round trip (opening and closing one position), with each position equal to the whole account and no leverage. This is not an Ouinex rate. Real costs depend on the market, the order type and how long a position is held; for actual charges, see the costs the ratio has to absorb.
Strategy A completes 12 round trips a month, so it pays 12 x 0.05% = 0.60% a month. Strategy B completes 2, so it pays 0.10% a month. Because the same cost comes off every month, the standard deviation does not change. Only the average falls.
Illustrative figures | A gross | A net | B gross | B net |
|---|---|---|---|---|
Round trips a month | 12 | 12 | 2 | 2 |
Cost a month (%) | 0 | 0.60 | 0 | 0.10 |
Average monthly return (%) | 1.20 | 0.60 | 1.30 | 1.20 |
Average excess return (%) | 0.87 | 0.27 | 0.97 | 0.87 |
Standard deviation (%) | 1.07 | 1.07 | 2.76 | 2.76 |
Monthly Sharpe ratio | 0.8079 | 0.2486 | 0.3505 | 0.3142 |
Annualised Sharpe ratio | 2.80 | 0.86 | 1.21 | 1.09 |
Rank | 1st | 2nd | 2nd | 1st |
Illustrative only. Hypothetical strategies and an assumed cost of 0.05% per round trip, not an Ouinex rate. Risk-free rate 4.00% a year (Bank Rate, 5 November 2026).
The ranking flips. Costs take about 69% of A's average excess return (0.60 of 0.87) but only about 10% of B's (0.10 of 0.97). A's steadiness was real, but its edge was thin, and frequent trading wore it down.
The cost level decides where the flip happens. Under the same assumptions, it comes at about 0.044% per round trip:
Assumed cost per round trip | A, annualised Sharpe ratio | B, annualised Sharpe ratio |
|---|---|---|
0.00% | 2.80 | 1.21 |
0.01% | 2.41 | 1.19 |
0.02% | 2.02 | 1.16 |
0.03% | 1.64 | 1.14 |
0.05% | 0.86 | 1.09 |
0.08% | -0.30 | 1.01 |
Illustrative only. Same hypothetical returns and trade counts as above.
At 0.08% per round trip, A's ratio turns negative: its returns no longer cover the risk-free rate. B barely moves across the whole table. The lesson is about trade frequency, not about one style being better than another. A backtest that leaves out costs flatters the strategies that trade the most.
Real costs are less tidy than a flat rate. Spreads and slippage tend to widen in fast markets, and overnight positions can bring financing charges that this example does not model.
What is a good Sharpe ratio?
There is no universal threshold: a Sharpe ratio is only good or bad compared with other strategies measured over the same period, with the same risk-free rate and the same cost assumptions. A single number without its window, its data frequency and its cost model tells you very little.
For perspective, Sharpe's 1994 article takes typical estimates for a broad developed stock market of a 6% annual excess return and a 15% standard deviation, which gives an annual Sharpe ratio of 0.40. Those are illustrative estimates, not a measured result, but they show why strategy A's gross 2.80 deserves suspicion.
Sample size matters as much as the number. The same article notes that the historic Sharpe ratio multiplied by the square root of the number of returns gives the t-statistic of the average excess return. With 12 monthly returns, that is exactly the annualised figure. So the net results, 0.86 for A and 1.09 for B, put each average only about one standard error above zero. Twelve months of data cannot tell either strategy apart from luck.
Andrew Lo's study of the ratio's statistics makes the wider point: Sharpe ratios are estimates with their own error, and in his hedge fund example, serial correlation overstated one fund's annual ratio by as much as 65%. A Sharpe ratio describes past risk-adjusted returns. It does not predict future ones.
Where the Sharpe ratio misleads
It assumes a bell-shaped spread of returns
Standard deviation fully describes risk only when returns are roughly normally distributed. One study of 2,763 hedge funds opens with the point that the Sharpe ratio is adequate for evaluating funds when their returns are normally distributed. Trading returns often have fat tails: rare, large losses that a bell curve treats as almost impossible. A strategy with small, regular gains and an occasional heavy loss can show a high ratio until the loss arrives.
It counts good surprises as risk
Standard deviation treats a large gain and a large loss alike. A CFA Institute article on downside risk notes that it penalises upside and downside deviations equally.
Autocorrelation breaks the square-root-of-12 rule
The annualisation in step 4 assumes each month is independent of the last. Lo shows that monthly ratios cannot simply be multiplied by the square root of 12 except in special cases. If good months tend to follow good months, the simple rule can misstate the annual figure, and correcting it can reorder fund rankings.
Smoothing can inflate it
Anything that makes reported returns smoother than real ones lowers the standard deviation and lifts the ratio. Research on hedge funds found that returns are often highly serially correlated, most likely because of holdings that are not actively traded, and that this understates volatility and increases risk-adjusted measures such as the Sharpe ratio.
It says nothing about the size of a loss
Sharpe shows that scaling a position up or down changes the average and the standard deviation by the same proportion, so the ratio is scale independent. A position run with leverage can have the same Sharpe ratio as an unleveraged one and far larger losses in a bad month. Trading leveraged products carries a high risk of loss.
Sortino and Calmar: two alternatives
The Sortino ratio keeps the shape of the Sharpe ratio but changes both parts. As the CFA Institute article describes it, a minimum accepted return replaces the risk-free rate on top, and downside deviation replaces standard deviation underneath. Only months below the target count as risk.
The Calmar ratio divides the annual return by the maximum drawdown, the largest fall from a peak to a later low. A study comparing hedge fund measures lists Calmar alongside Sharpe and Sortino among the ten measures most used by researchers and practitioners.
Run on the net returns of the two hypothetical strategies, they disagree:
Illustrative figures, net of the assumed costs | Strategy A | Strategy B |
|---|---|---|
Annualised Sharpe ratio | 0.86 | 1.09 |
Annualised Sortino ratio (target 0.33% a month) | 1.48 | 2.07 |
Maximum drawdown (%) | 1.20 | 3.50 |
Calmar ratio (12-month return / maximum drawdown) | 6.15 | 4.26 |
Illustrative only. Hypothetical strategies. Downside deviation here counts months above the target as zero and divides by all 12 months.
Sortino agrees with Sharpe and ranks B first. Calmar ranks A first, because A's worst fall was small. Twelve months hold only one or two drawdowns, so a Calmar figure from so short a window rests on very little data. Downside deviation has its own trap: the CFA Institute article warns that using only the historical returns below the target can significantly underestimate downside risk.
On large samples the measures often agree. The 2,763-fund study found virtually identical rank ordering across 13 measures. On short samples, with very different trade frequencies, they can split, as they do here. Read them together.
How algo traders use the Sharpe ratio to compare backtests
An algorithmic strategy is a set of written rules, so it can be tested on past data before it runs. The Sharpe ratio is a common yardstick for those tests because it puts strategies of different sizes on one scale. Used well, it follows a short checklist:
- Net of costs, with realistic trade counts. As the worked example shows, the gross figure can rank strategies in the wrong order.
- Same window, same frequency, same risk-free rate for every strategy in the comparison.
- Count the variations tried. Researchers writing on backtest overfitting found that the higher the number of configurations tried, the greater the probability that the backtest is overfit, and that most backtests do not report that number.
- Keep some data back. Test the chosen settings on a period the rules were not tuned on, and expect the ratio to fall.
- Look beyond one ratio. Check drawdown, Sortino and whether monthly returns are correlated before trusting an annualised figure.
The aim is a fair comparison, not the highest number. Pair it with sound risk management, and if you build rule-based strategies, you can run a strategy on Ouinex.
FAQ: Sharpe ratio questions answered
What does a Sharpe ratio of 1 mean?
An annualised Sharpe ratio of 1 means that, over the period measured, the average return above the risk-free rate equalled one standard deviation of those excess returns, on a yearly basis. Whether that is high or low depends on the comparison: the same period, the same risk-free rate and the same cost assumptions.
Can a Sharpe ratio be negative?
Yes. A negative Sharpe ratio means the strategy's average return was below the risk-free rate over the period. In the worked example, strategy A turns negative once trading costs reach 0.08% per round trip. Comparing two negative ratios needs care, because more volatility then makes the figure look less bad.
What is the difference between the Sharpe and Sortino ratios?
The Sharpe ratio divides excess return by the standard deviation of all returns, so large gains count as risk. The Sortino ratio divides by downside deviation, which counts only returns below a chosen target. A strategy with occasional large gains usually scores better on Sortino than on Sharpe.
Why multiply by the square root of 12?
To turn a monthly Sharpe ratio into an annual one. If monthly returns are independent, the average grows twelvefold over a year while the standard deviation grows by the square root of 12, so the ratio grows by the square root of 12. When returns are correlated from month to month, this shortcut can misstate the result.
Conclusion
The Sharpe ratio is a simple division: average excess return over its standard deviation. The worked example shows why the inputs matter more than the formula. Before costs, the frequent trader scored 2.80 against 1.21. After an illustrative 0.05% per round trip, it scored 0.86 against 1.09, and the ranking reversed.
Compare strategies net of costs, over the same window, with the same risk-free rate, and treat any figure from a short or heavily tuned backtest with caution. A Sharpe ratio describes past risk-adjusted returns and does not predict future ones.
Trading leveraged products carries a high risk of loss. See how algo trading works on Ouinex.
Sources
- William F. Sharpe: The Sharpe Ratio, Journal of Portfolio Management, Fall 1994, reprinted on Sharpe's Stanford page. Origin in Sharpe (1966) "Mutual Fund Performance", Journal of Business, January 1966, pp. 119 to 138; the reward-to-variability name; riskless benchmark; AVERAGE / STDEV method; annualisation under zero serial correlation; t-statistic relation; scale independence; the 0.40 stock market illustration.
- Bank of England: Interest rates and Bank Rate, read 17 November 2026. Bank Rate raised to 4%, published 5 November 2026.
- Bank of England Database: Official Bank Rate history, read 17 November 2026. 3.75% from 18 December 2025 to 5 November 2026.
- Andrew W. Lo: The Statistics of Sharpe Ratios, Financial Analysts Journal, vol. 58, no. 4, July 2002. Annualising by the square root of 12, estimation error, the 65% overstatement example.
- Mila Getmansky, Andrew W. Lo and Igor Makarov: An Econometric Model of Serial Correlation and Illiquidity in Hedge Fund Returns, NBER Working Paper 9571, March 2003. Smoothing, serial correlation and inflated Sharpe ratios.
- Deborah Kidd, CFA Institute: The Sortino Ratio: Is Downside Risk the Only Risk that Matters?, 2012. Sortino formula, downside deviation, limits of standard deviation.
- Thi Thanh Huyen Nguyen: Assessing Hedge Fund Performance: Does the Choice of Measures Matter?, working paper, October 2007. Calmar among the ten most used measures.
- Martin Eling and Frank Schuhmacher: Does the choice of performance measure influence the evaluation of hedge funds?, Journal of Banking and Finance, vol. 31, no. 9, 2007. Normal-distribution condition; rank ordering across 13 measures for 2,763 funds.
- David H. Bailey, Jonathan Borwein, Marcos López de Prado and Qiji Jim Zhu: Pseudo-Mathematics and Financial Charlatanism: The Effects of Backtest Overfitting on Out-of-Sample Performance, Notices of the American Mathematical Society, May 2014. Backtest overfitting and the number of configurations tried.



