Say you are designing a system that looks for low vol and then a rapid expansion... pretty vanilla stuff.
I believe that testing such a system on back-adjusted data does not work. Make a series that rolls on OI. The spot contract will stay in place until OI drops below that of the next contract in the series. In many markets, often short dates i-rates, the volatility will drop off into expiry as OI falls. Then the new contract is appended to the data and that new contract is displaying 'normal' volatility. If you look at the back-adjusted series it will appear as though the market went very quite and then had a big volatility expansion and this point lines up very nicely with the roll date from one contract to the next (in the back-adjusted series). This, to me, is a fake volatility expansion.
Someone must have dealt with this?
Is there a back-adjusting method that accounts for differences in volatility?
thanks yet again for any feedback.
damian
Volatility expansion - test on back-adjusted data
-
- Roundtable Knight
- Posts: 118
- Joined: Tue Apr 15, 2003 7:44 pm
- Location: Arizona
Backadjusted data series is "wrong"?
The key insight is to figure out how you, the human trader, intend to roll over positions from one contract to another when you're trading the system live, using real money. Think, papertrade, ruminate, fiddle with charts, talk to friends, visit a psychiatrist, read books, ask your broker, employ your process whatever it is. Decide exactly how you want to rollover for this specific system. Do this first. Then configure your data software to do it that way. Voila, a perfect match .
There may be two problems: (1) after thinking long and hard you may decide that your rollovers for this system will employ human judgement. Congratulations, you have discovered you've got a discretionary system, and good luck testing it. (2) You may have chosen a 100% purely mechanical method for rolling over, but your data software doesn't support it. In that case you'll need to build or buy new software, abandon the system idea, or think it over some more. Maybe you'll find a method of rolling over that you are happy to trade, and which your data software happens to support.
There may be two problems: (1) after thinking long and hard you may decide that your rollovers for this system will employ human judgement. Congratulations, you have discovered you've got a discretionary system, and good luck testing it. (2) You may have chosen a 100% purely mechanical method for rolling over, but your data software doesn't support it. In that case you'll need to build or buy new software, abandon the system idea, or think it over some more. Maybe you'll find a method of rolling over that you are happy to trade, and which your data software happens to support.