How does one professionally trade the vol surface assuming they have direct market access to several exchanges (we're talking paying port/session fees) with latencies in sub millisecond and book data for both underlying AND contract series. I'm just curious as to what general concepts are used and how much revenue one can make doing this on US equities and the S&P 500 product suite.
I'd love to hear how you can set that up as a Customer or Pro-Customer with those latencies. I don't expect that is possible even as a Member/MM without a very expensive hosted setup.
Not possible... trading vol surface is basically a MM strategy, where they buy the cheap outliers and sell the expensive outliers. Risk is mitigated by having positions and trading across the board. So unless you have very good fee structure including decent lending facilities, and can compete on speed with all other MM's and hedgefunds that have algo's constantly watching the entire surface... it's not going to be work... especially on S&P500, which is probably the most liquid and efficient market on the planet. You might get away with illiquid stocks and options, but then you have to accept a greater risk, since that lack of liquidity also means you can't build up and reduce your options book at anytime.
see if you can get a trial of Orc Trader or Option City metro (for example) and see what it is they allow you to do. Then do that...(which is probably pretty much in line with JackRab's suggestion)
I have a pretty extensive spreadsheet system that plots spread revenue against position delta (as, mostly, vol differences are *way* too fuzzy). That system directs me to -- guess what! -- sell high (vol) and buy low (vol). This is sub-hour stuff, mostly. Sometimes sub-day. A boat needs water to be a boat, and don't behave so well in froth. In an ocean of nickels (and sometimes, quarters and fiddy-cents), worrying about sub-pennies' spray won't get you to shore. [sorry. I had a twisted youth.]
Basically anything that's not according to your vol surface... bigger deviations are better of course.