I'll give you something I think is worth trying if you promise to let me know how it goes, ok? :)

Deal

Cool. These funds do shorting allowing for us to go long on them, sort of mirroring SPY, can see that at http://finance.google.com/ entering spy and then adding these in the 'Compare' box right above the chart. They can also lower beta (contest).

context.hedge_shorting = symbols('RWM', 'HDGE', 'SEF', 'DOG')

The reason I'm so interested is that I've had various results with them and haven't quite taken the time to pin down why.

Edit: Maybe those can come in handy for someone else (having seen the message below now).

Thanks for the suggestion, but Im looking for something that would go with the market, not against it. Basically, shorting the VXX is assuming a reduction in volatility, which usually happens when the market goes up.

If it works with VXX but fails with XIV then you are not betting on volatility going down, you are betting on something else.

The system doesn't actually fail, but it is not nearly as profitable. The reason is that VXX and XIV do not have a 100% inverse correlation.

I'm not looking for 100% inverse correlation - I understand that doesn't exist - I'm just looking for some ideas of other things I could try.

Yes, well, feel free to try VIXY SVXY UVXY etc, but I think you'll find that your system requires shorting to be profitable.

Did you look at when these funds take their fees or resize their positions? That might explain the difference if you're only holding overnight.

Did you also check on the short availability. Maybe the effect you've found exists because it's not possible to consistently short vxx overnight. Or the spread is huge, much larger than Q's estimates.

Quantopian doesn't have options, but I'm sure your strategy could be implemented in a limited risk limited reward manner using an option strategy.

Short TVIX and long VXX. Equally weighted example.