Why are overnight periods riskier? For one, you can’t use stops to limit your risk. But more importantly, the distribution of overnight returns has far more extreme negative returns than the intraday or close-to-close periods. Let’s take a look at some stats on close-to-open, open-to-close, and close-to-close returns for SPY:
- Skew: negative skew means longer tails on the negative side.
- Kurtosis: higher kurtosis means heavier tails (the normal distribution has a kurtosis of 3).
First of all, overnight returns are not that volatile. But per-unit-volatility, they are far riskier due to the higher frequency of extreme returns. I calculated the magnitude of returns in terms of standard deviations (based on 10-day realized volatility), and overnight returns have a 5 standard deviation move more than twice as frequently as close-to-close returns. You can expect a 5SD overnight move about once a year.
Here’s a histogram of close-to-close returns and close-to-open returns. You can clearly spot the skewness around the -6% to -3% area. The value on the far left is from Oct 24 2008. Also note that just because the skewness is negative doesn’t mean that short positions are safe: the tails of the overnight returns are heavy on the right side as well.
These risks are about the same in bull and bear markets. On the other hand they seem to be smaller in high-volatility environments (possibly a side-effect of mean-reverting volatility). When SPY’s 20-day realized volatility is above 20%, the tail risks of overnight returns are about the same as those of close-to-close returns. But this isn’t all that helpful, because after all in times of high volatility your position sizing is already limited.
The weekend is a special case of even higher risk; you can’t treat it like other overnight periods. It’s more volatile, and features even more frequent extreme losses. In my own position sizing I always put additional limits on trades over the weekend.
Different instruments behave in different ways, especially when it comes to different asset classes. Overnight dangers tend to be relatively larger for stocks compared to indices. So make sure that your sizing is adapted to the unique characteristics of the instruments that you’re trading.
As a rule of thumb I’d say that given an equal volatility exposure, overnight returns expose you to about ~1.5-2x risk of extreme negative returns, while the weekend exposes you to 2-2.5x risk of extreme negative returns. But does this mean that you should halve any overnight trades that you enter on Fridays? Not necessarily. There is no perfect answer to this dilemma: different traders and different strategies will have different views on how much they are willing to trade returns for less extreme moves. The important part is to quantify the trade-off so that you know exactly how much you’re giving up and how much risk you’re taking on.
Quantocracy's Daily Wrap for 11/09/2017 | Quantocracy says:
[…] Overnight Risk [Qusma] […]