macrocephalopod

To model this you should build constant-maturity VIX curves by interpolating forward VIX prices to a floating grid of expiry terms (e.g. 1 mo, 2 mo, 3 mo) rather than a fixed grid of expiry dates. You then build a model to forecast the changes in constant-maturity forward VIX

for whatever forecast horizon you care about, and combine it with the observable rolldown to get a forecast for the P&L for particular points on the constant-maturity curve, which you can then interpolate back to a forecast for the actual contracts.

For obvious reasons I won't elaborate on how you should forecast price moves, but I will note that (a) the term structure itself is important and (b) constant-maturity VIX prices are highly mean reverting. Add your own intuitions on top of that!

My final point is that nothing I've said here is particularly specific to VIX, but VIX is a very clean example because the underlying is not tradable (so there aren't any arbitrage relations and everything is priced on expectation) and it's not seasonal.

With suitable adjustments to take arbitrage bounds and seasonality into account, this is a very general framework for trading all kinds of futures. FIN.

Gonna tag the vol nerd crew for their input @KrisAbdelmessih @Ksidiii @bennpeifert @pauleluard @SqueezeMetrics @NewRiverInvest @VixCentral @VolQuant @QuantVol @varianceswap @selling_theta

I get that vol is sexier than other derivs, but don't understand why people want to start there first. Term structure and PnL by source of return took me a while to grasp as a junior (though to be fair I'm pretty stupid). Think people should start with Delta One products first.