Semideviation is the volatility of a strategy's losing days alone: the standard deviation of the daily returns below zero, annualized with the square root of 252 like volatility itself. It measures how much the losing days differed from one another — whether the bad days were all alike or ranged from mild to much worse — and leaves the winning days out entirely.
Also seen as: semi-deviation, semi-standard deviation, downside volatility
How is semideviation calculated?
Keep only the days the strategy lost money, take the standard deviation of those returns around their own average, and annualize it:
where 252 is the number of trading days in a year, the same convention volatility uses. In words: it is volatility, measured on the losing days only. A strategy whose losing days were all about −0.5% has a low semideviation; one whose losing days ranged from −0.1% to −4% has a high one.
Why is semideviation about 0.6 times volatility even for symmetric returns?
Because it measures the losing days around their own average, not around zero, semideviation is not the share of the movement that went down. Take returns that are perfectly symmetric — as many ups as downs, of the same sizes: their losing half, measured around its own average, has a standard deviation of about 0.6 times the full volatility. So a semideviation near 0.6 × volatility is the ordinary case, not a sign that most of the movement was upward; a figure well above it means unusually uneven losing days.
How does Fincanva handle it?
- Performance Metrics shows it in the Volatility and risk group as Semideviation (downside only), right under Volatility, annualized the same way — see what every number in Performance Metrics means.
- It follows the same losing-days convention as the Sortino ratio on the Components tab.
- A simulation computed before the row existed shows "n/a", never 0%.
What does it look like in practice?
Two strategies both show a Volatility of 15%. The first reads Semideviation (downside only) 9%, close to 0.6 × 15%: its losing days were about as varied as a symmetric distribution expects. The second reads 13%: its losing days ranged much more widely — most mild, a few far worse — even though its overall volatility is the same.
What counts as a good value?
Lower means more uniform losing days. Compare it with the strategy's own volatility first: about 0.6 × volatility is ordinary, well above means some losing days were much worse than the rest, which is worth reading beside the Value at Risk and Conditional Value at Risk. Between strategies, compare it only over the same period.
Where this term is used
Auto-generated · 1 pageThe pages that use this term: read it in context there.