I'm trying to implement a few simple VIX strategies (1/0/-1-signals based on MA crossover, term structure, hvol vs ivol) in Python. I am new to quant and volatility, but looking at the VIX properties suggets that a short signal bias (-1) should be successful (I multiply the signals with the log returns of daily closing prices of the 1-month VIX futures). These strategies are not successful and my guess is that I am not correctly exploiting the longer term properties of the VIX. Instead of going long/short daily, is it possible use weekly/monthly closing prices eg. ln(Price t/Price t-30) * (-1), and the benchmark would remain at ln(Price t/Price t-1)?
This is not an answer, but instead advice:
Since you're new to quant and volatility then you should start with something other than a volatility or a rates product because those are going to be some of the most complicated products. First, you need to gain some comfort with quant stuff. Then, you can move onto more complicated topics. I'm not saying that you can't do it, I'm just saying that you shouldn't even try at this stage because you're more likely to hurt yourself than help yourself-- because the number of mistakes you could be making is basically infinite. Find some simple models for just about anything online and try to reproduce them from scratch (not by just copy-pasting).
If you're deadset on looking at this, as a first step try to replicate the returns of some of the ETNs in the space. This will help you find a number of errors in your code. Second, remember that if you find anything in any product, keep it to yourself.