Beef satay, chicken satay, stingray, marshmallows, lollipops, sour plums covered in caramelised sugar (worth trying in Beijing) not to mention pancake-sausage-on-a-stick.
Take a piece of food, skew it with a stick and stick it in your mouth.
The skewness statistic measures whether returns are more limited on the upside or downside.
Many strategies can be viewed from a skew (or gamma and convexity) search perspective (or a negative skew minimising one).
For example, value investing can be seen as picking stocks whose price has bottomed out. Limited downside is secured by good accounting ratios on the one hand, higher upside is seen in lower market prices on the other.
Growth stock picking is similar; but a very large upside is the big story. So humongous in fact that the downside needs only be limited by zero and the growth story is still a good story.
Throw momentum and mean reversion onto that list.
Buy Low Sell High
So how about using skew as a buy or sell indicator?
Assuming that skew is somewhat sticky; let's go long for a day if the last week's skew was positive and vice versa.
On 65 year's worth of S&P returns. The result is a Sharpe of about 0.1. Meaning there's a rather high one in five chance that strategy won't be profitable over the long term.
Not a good result.
Especially when I haven't factored in trading fees; plus the nature of the beast probably means a lot of those!
Nevertheless, might be worth some more investigation.
Two things are of interest.
Are statistical skewness and profitability actually linked (it's a lot weaker than gamma and convexity) if so can we come up with a predictor of skewness?
Skewness from historical returns do not seem to be promising. Perhaps the implied skew index would be better?