One can trade vol swap to get exposure of the volatility of the underlying security in a 'clean' way. On the other hand, we know that vol swap, theoretically can be replicated by a dynamic position of options. However, it is not practical to trade the whole chain of options. The payoff of a single dynamically hedged option of a particular strike is dependent on the path of the underlying security, and therefore not a 'pure vol' trade. If one trade dynamically hedged position of a straddle/strangle, then it depends less on the path. My question, if one would like to trade vol with options, in practice, what is the best way so that we can have a somehow 'clean' exposure to the volatility?
To trade forward starting volatility swaps, see this paper (actually more a practitioner's scribble than a paper) and all references therein: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3354408
Just to let you know, I think there may be a small error term in the paper that needs to be included still, but based on some simple numerical tests I was able to get a hedge accuracy of 99% or more, without this small additional missing term, by trading zero vanna forward starting straddles with a specific skew dependent notional for every 0.5-1 volpoint move in implied vol, which is not too bad; better than dynamically trading a whole strip of options. I'll update the paper in due course with that small error term I mentioned. The note will tell you how to choose the notional appropriately.
I am still working on a same kind of strategy for trading realized vol, will post it online when ready. So if you're looking to trade realized volatility I guess for now there are only 2 options (AFAIK):
- Follow Carr-Lee --> dynamically trade strip of options, this is model-free for the class of stochastic volatility models specified in their paper
- Assume log-normal model for realized volatility (or some other model if you wish) and dynamically hedge realized volatility using variance swaps (but varswaps are a strip of options, so this is comparable to #1 above, but much more model dependent)
[PS I'm not a fan of referencing one's own work, but I really don't know (m)any other papers that treat volswap hegding/replication using only a straddle].