The Mojito uses a step function to switch up allocations between the VXX and VXZ ETFs over time.

A bunch of rules which says...

If the spot VIX to VXV (3 month VIX future) ratio ('IVTS') is lower than 0.91 then short the VXX (short term VIX future ETF) and weight by -0.7, while long the VXZ by 0.3;

and so on...

Each dot on these graphs relates to those rules, imagine a straight line connecting them together.

I noted previously that the strategy lost some of its zest after 2012; and wondered whether the 'steps' were overfit to pre 2012 data in order to gin up performance.

So, I fit a line line to the steps in order to make it a little more generalised. Perhaps if there was some sort of more general rule behind the steps, it could be extracted into a continuous function.

The 'generalised' version of the Mojito does well (1.25 Sharpe) but its Sharpe Trajectory is still less efficient in patches (original has a Sharpe of 1.6).

Oddly it seems to generate more excess volatility during good times when it's also earning money - reminding me of the reason why this strategy beats Equal Weighting.

The gap narrows a tad after 2013 (1.3 vs 1.1) but not enough.

While these step functions are prone to being ginned up; this particular one looks quite robust.

Another big plus to the Mojito is that the post 2013 performance is still very good - if not as stratospheric as before - everything is relative as they say.

It pays to put things in context before admonishing, and shows us that the original Mojito was pretty darn smooth even after 2013!

Having said all that this is all based on only 5 years of data, more would be nice!

Code for the Lazy Backtesting IDE is here.