OptiNod Academy

Sharpe and Sortino: Performance Divided by Volatility, and the Problem With Penalizing Upside

Comparing strategies by return alone ignores how differently they may behave on the way to the same result. This article explains the limits of the Sharpe ratio and why the Sortino ratio is often closer to real trading conditions.

> Two strategies can reach the same return, but when the path to that return is more volatile, one may be much harder to hold in live trading. The Sharpe ratio measures that path, but it also treats upside surges as risk.

The Sharpe ratio, formalized by William Sharpe in 1966, is a simple ratio. It subtracts the risk-free rate from a strategy's average return, then divides the result by the standard deviation of returns. The numerator is the excess return earned for taking risk. The denominator is how much returns fluctuated while earning it. In one number, Sharpe compresses how much excess return the strategy generated per unit of volatility.

Most traders compare strategies with a single return figure. If Strategy A returns 80% per year and Strategy B returns 60%, they conclude that A is better. That comparison ignores the denominator. If A's equity curve swung 30% on the way to 80%, while B's moved only 12%, few traders will stick with A to the end in real trading. They exit or cut size during the drawdown. Two strategies with the same return but different volatility are completely different strategies.

Sharpe itself has a structural weakness. Standard deviation in the denominator counts upside and downside moves equally as risk. A day that jumps +10% increases volatility and lowers the score. Traders are usually afraid of downside, but Sharpe treats welcome upside spikes as risk too. The Sortino ratio addresses this weakness by putting only downside volatility in the denominator.

Same Return, Different Volatility, Different Strategy

Sharpe starts from a simple idea: divide return by volatility. A strategy that produces the same excess return with lower volatility receives a higher score. In practice, this denominator matters because traders have limits to how much drawdown they can tolerate. A highly volatile equity curve creates deep troughs along the way, and most traders liquidate or reduce exposure in those troughs. If you cannot hold the curve to the end, the final return is never realized.

BTC gives a useful real-world example. In Q1 2023, from January 1 to April 1, BTC rose from $16,617 to $28,453, a gain of about 71%. Over the same period, the standard deviation of daily returns was about 2.96%. If the same 71% return had come with daily volatility in the 1% range, the equity curve would have been much smoother, and the psychological pressure on the trader would have been completely different. A single return number hides that difference entirely.

The risk-free rate in the numerator is also worth noting. Sharpe uses excess return: total return minus the risk-free rate. The risk-free rate is what the same capital could have earned without taking risk, usually represented by short-term government bond yields or money market rates. Depending on the period, it may be close to 0% or around 4-5% annually, so it is not a fixed value. For crypto strategies with annual returns in the tens of percent, setting the risk-free rate to zero may not change the result much, but comparisons must use the same risk-free rate or the numerator will be inconsistent.

All of these calculations begin with the same question: how much did the strategy earn per unit of volatility? The return itself comes next. A strategy that makes 50% per year with 25% volatility may scale better at the same leverage than one that makes 80% with 60% volatility. Lower volatility leaves more room to increase position size within the same risk budget. The higher-return strategy is not always the better strategy.

Sharpe Counts Upside Moves as Risk Too

The denominator of the Sharpe ratio, standard deviation, squares and adds every deviation from the mean. It only measures distance, whether the move is above or below the mean. This creates a contradiction. Traders experience downside moves as losses, but Sharpe includes a welcome +10% day in volatility and penalizes the score. The more often a strategy has large upside moves, the larger the denominator becomes and the lower its Sharpe ratio appears.

This effect is especially important in crypto. BTC often has upside spikes as large as, or larger than, downside selloffs. In Q4 2024, from October 1 through year-end, BTC's largest positive daily return was +10.3%, while its largest negative daily return was -5.59%. The biggest move was to the upside. Sharpe counts that +10.3% as volatility, increasing the denominator and lowering the strategy's score. The trader's best day becomes a penalty in the metric.

Recalculate the same period with Sortino and the picture changes. In Q4 2024, BTC's daily standard deviation was about 2.53%, but its downside deviation was about 1.31%. With the denominator cut roughly in half, Sortino came out about 1.93 times higher than Sharpe. Same return, same data, but once upside surges are removed from the risk calculation, the strategy's evaluation nearly doubles. The larger this gap, the more upside-heavy the strategy is.

Sortino Uses Only Downside Deviation

The Sortino ratio keeps the numerator and changes only the denominator. It selects only returns below the target return, usually 0, squares and adds them, and treats days above the target as zero in the denominator calculation. This is downside deviation. Upside volatility is not counted as risk, so the denominator reflects only the loss volatility the trader actually has to endure. For the same strategy, the larger the upside volatility, the wider the gap between Sharpe and Sortino.

BTC in Q1 2023 shows this gap clearly. Daily standard deviation during the period was about 2.96%, while downside deviation was about 1.44%, less than half. Across the 90 trading days, the number of up days and down days was similar, but the largest gain, +9.62%, was bigger than the largest loss, -6.19%. Large moves were concentrated on the upside. As a result, annualized Sharpe was about 4.1, while annualized Sortino was about 8.5, more than twice as high.

How you read that gap matters in live strategy selection. If Sortino is much higher than Sharpe, most of the strategy's volatility comes from the upside. Trend-following strategies often look like this. They cut losses often and keep them small, while taking large profits in a single move, so the big upside moves drag Sharpe down. If you reject such a strategy by looking only at Sharpe, you may miss one of the easier strategy profiles to trade. Sortino shows that profile more accurately.

Do Not Use Stock Annualization for Crypto

Sharpe and Sortino are usually compared after annualization. The usual method multiplies the daily metric by the square root of the number of trading days, but the number of trading days differs by asset class. Stocks have about 252 trading days per year, so standard deviation is multiplied by √252, or about 15.9. Crypto trades 24/7, including weekends and holidays, so the correct factor is √365, or about 19.1. Since the ratio between the two factors is about 1.20, using the stock factor for crypto understates volatility by about 17%.

That difference distorts comparisons. Annualizing BTC's Q1 2023 daily standard deviation of 2.96% with √365 gives about 56.6%. If you mistakenly use √252, it becomes about 47.0%. That understates volatility by 9.6 percentage points, inflating the Sharpe ratio because the denominator is too small. This is why stock-oriented backtest tools can make crypto Sharpe ratios look better than they really are when applied directly to crypto data.

This may seem like a simple matter of choosing the right annualization factor, but it is especially dangerous when comparing asset classes. If you compare a stock strategy and a crypto strategy side by side and one uses 252 while the other uses 365, the comparison is misaligned. Always check which asset class the backtest's annualization factor assumes. A strong-looking Sharpe may be an artifact of the calculation method.

Sharpe Underestimates Risk in Fat Tails

Sharpe and Sortino share the same assumption: they measure volatility with standard deviation. Standard deviation summarizes risk well when returns are close to normally distributed. Crypto returns, however, have fatter tails than a normal distribution. Markets may stay quiet most of the time, then occasionally suffer selloffs so large they should be nearly impossible under a normal-distribution view. Standard deviation does not fully capture the weight of these rare large losses.

Sharpe rises during long calm periods. When a low-volatility sideways market persists, the denominator shrinks and Sharpe improves. But that same calm may be hiding the risk of the next major selloff. This was the pattern in the months before LUNA collapsed in 2022 and before several altcoins were swept into cascading liquidations. Assets that had maintained low daily volatility fell by tens of percent in just a few days, and the prior Sharpe ratio gave no meaningful warning of the loss magnitude.

Sortino reflects this risk somewhat better than Sharpe because it looks only at downside moves, but it still has the same standard-deviation limitation. Both metrics are more sensitive to frequent small moves than to a single major loss. For fat-tailed assets, Sharpe and Sortino should be read alongside maximum drawdown, or MDD. If you do not also check the deepest trough in the equity curve, a high Sharpe may simply come from a period that happened to avoid a major crash.

Short Samples Make Both Metrics Unreliable

Sharpe and Sortino are calculated from averages and standard deviations, so both values can swing widely when the sample is small. A Sharpe of 1.5 from two years of data is far more credible than a Sharpe of 3.0 from a 30-trading-day backtest. In a short window, volatility may have been low by chance or the trend may have run in one direction. That luck shrinks the denominator and inflates Sharpe. A high Sharpe from a short sample should usually be treated with little confidence.

The problem is worse for Sortino. Its denominator uses only returns from down days, so if a short sample has only a few down days, downside deviation becomes unstable. In a window that happens to include very few losing days, Sortino can become unrealistically high. The distortion is larger in short, strong uptrends. Down days are rare, the denominator gets small, and Sortino reports a value that is difficult to reproduce in live trading.

Use this checklist when validating the metrics.

• [ ] Sample size: Is the backtest at least one year long, or at least 250 daily bars? If not, treat Sharpe and Sortino as reference values only.
• [ ] Annualization factor: Did you use √365 for crypto and √252 for stocks? When comparing asset classes, confirm that the factors are aligned.
• [ ] Number of down days: Are there enough down days to form Sortino's denominator? If there are fewer than 20 down days, do not use Sortino as a standalone basis for a decision.
• [ ] Maximum drawdown check: Did you also check MDD for the same period? High Sharpe combined with deep MDD is a warning sign for fat-tail risk.

Read the Two Metrics Through Their Gap

Sharpe and Sortino work together as a pair. They are most informative when read together through the gap between them. If Sortino is much higher than Sharpe, the strategy's volatility mainly comes from the upside. Strategies that capture large gains at once, such as trend following, often fit this profile. If the two values are similar, upside and downside volatility are close to symmetric, meaning the strategy has no strong directional bias in its volatility.

By contrast, if Sortino is only slightly higher than Sharpe, or nearly the same, and the absolute values are low, loss-side volatility is as frequent and severe as profit-side volatility. Even with the same average return, this kind of strategy is hard to hold in live trading because the equity curve has deep troughs. As with BTC in Q4 2024, periods where Sortino is nearly twice Sharpe are upside-heavy markets, while periods with a narrow gap have more evenly distributed two-way volatility. The gap changes over time even for the same asset.

When choosing strategies in practice, resist the impulse to look at return first. Start with performance per unit of volatility. Then use the gap between Sharpe and Sortino to identify which direction that volatility came from. Finally, check maximum drawdown and sample size to judge whether the number is repeatable or just noise. These three steps reveal the real shape of a strategy that a single return figure hides. The highest-return strategy is rarely the easiest one to scale, far more rarely than most traders assume.