I guess I stand corrected - just went through the math in the prospectus and it does look like roll amount CRW is weighted weighted by the notional value of the futures. This way you get roll proportional to the ratio of the futures, not to the absolute difference. Although at these levels it’s an insignificant difference, it could make for interesting behavior sometimes. For example, imagine how quick vxx recovers value when the curve is inverted by 45% ratio like it was in October of 08 - the reverse decay is like 2% a day
I'm referring to the really way OTM stuff. In those, what is not in the backtest data is even more important than what is in it. People will draw confidence from 10-15 year backtests, then one day their account gets liquidated by their broker
Spitznagel has a backtest in this book. It shows how an index position that is paired with some put buying (really way OTM puts) can outperform an index position by itself but only if the put buying is reserved to periods where the market is overvalued (by his definition)
I’d be very skeptical of any back tests and methods involving the tail events. A very small number of parameters will let you fit the history very well without much relevance to real life. In real life, you end up either with a bunch of false positives (and keep buying puts) or you end up with a false negative when it matters the most. In the worst case, you will end up with both, overspending on protection in normal times and yet getting hurt in the downturn.