More like pseudo dispersion. Here's my thought process: 1. a few big players contribute to most of the S&P 500's variance, say 35. 2. Those players are dynamic except for maybe the top 15 that are always on the list (this may not be true, but let's assume for a moment) 3. Run PCA (principal component analysis), figure out who those players are, then break down the factors to the most non-correlated sectors 4. sell vol on this basket, and adjust periodically to reflect the variance contributions by those players 5. buy vol on the index I'm going to run simulations, and see how it works out, if there's any edge or if transactions/slippage will kill me. Is there anything we need to add?? thanks
Ok. I am a little uncertain how you are going to structure you data so as to be able to tell who the institution ("players") in the market is at any given time? And why would that matter? Imo, this is poorly thought out. It is as if you took a bunch of words that people have used to analyse markets (PCA, Dispersion, Factors, etc), and threw them together haphazardly into sentence that resulted in a vague idea. At least for me, it is difficult to give advice in this context.
Well, let's say you run PCA and it gives you a set of factors with the factor loadings. How are you intending to establish the actual mkt significance of each of these factors? Why would you necessarily expect that a small number of significant factors that you will extract actually correspond to a similarly small basket of stocks?
By players I meant stocks...35 stocks contribute to vol the most, 15 are always on the list, break down to 5 components and trade accordingly. Buy vol on index, which should theoretically be cheaper.
And the index in this case is the SPX? I think this is a reasonably idea, but I think your universe should start at 500 stocks, and try to get to 50, not start at 50 and get to 15. Also, how will you be able to tell relative cheapness/expensiveness? Indentifying the stocks is one thing, but that is not even half the game.
Dispersion trading is a pretty popular strategy. The idea is that IV is over-priced with components of an index relative to the index itself. (I think about this by saying: investors with large exposure in specific stocks *must* hedge with options on that stock, making them more expensive than the broader index.) I don't really see a fundamental problem with what you're doing, nor do I have any insight whether it will be profitable. I do know that iVolatility actually sells some data specific designed for dispersion trading. See: http://www.ivolatility.com/eda.j
The thousands of firms who run dispersion off the SPX all have done their own correlations for the components and typically they all use somewhere in the neighborhood of 27 to 50 different components. There is really nothing new here and the dispersion book has been an overcrowded strategy for quite some time. Those who are successful at it generally make markets in many of the components and have access to the order flow in them in order to take advantage of the tiny edge. Dispersion is not realistically a strategy employed by individuals.
I expect that to be the case for most popular strategies--if not all strategies. I understand a lot of the edge is gone, but I'm willing to speculate on vol, instead of arb it away. I can't compete in the arb space, anyway. Isn't there always a better and more sophisticated trader out there?