Previously, I checked whether historical skewness was a good indicator to buy and sell the S&P 500.
My backtesting framework can now use the implied skew index as an indicator to buy or sell.
Now, the strategy buys the S&P 500 if the implied skew index has dropped day over day and vice versa.
Note that the S&P 500's implied skew is permanently negative, as huge drops are more likely than large jump in the short term.
Once I get a chance, I will make it easier to change horizons so I can test month over month (the skew index's horizon is 30 days).
The Sharpe has increased to ~0.25 and there is just a one in ten chance of the strategy being complete jibberish.
Interestingly the previous strategy now has the same Sharpe when tested over this shortened time period - the skew index only goes back to 1990 instead of 1950.
How many strategies are thrown away because they don't test well during Black Monday in the 80's for example? Or how many do something special to look good for that handful of days?
Not that it should be ignored entirely.
Another reason to use 'parallel universe' back testing.
The only advantage with using implied data right now is that we only need a single daily return to make a decision (versus a week's worth of data). Conserving data is a nice concept.
[Backtesting code is improving but rough. Download here]