Hmm, it's not that easy to backtest because you'd have to know the actual IV's of the time, not only to determine the price of the original straddle but also to know where the hedgepoints are every day. 2) I'm not sure what you mean by 'close intraday' and 3) if you mean the PnL of the adjusted strategy as a whole, so including the cost PnL of the straddle, I would agree. My opinion is that, unless you're very very good at predicting the stock, you're better of not adjusting at all. And we know that letting straddles expire will give you an average PnL of zero/0. Ursa..
LOL. Have no idea what I meant either. Essentially, if one were to hedge continuously a la Black Scholes then theoretical PnL would be zero OR if implied vols were below statistical vols then the realization/crystalization of that would be reflected in a positive PnL. However, continuous hedging is not possible and hedging at discrete intervals manifests a "replication error". The smaller the hedging interval or higher the frequency, the smaller the "replication error". It is desireable for this strategy to have a large positive "replication error" in order to overcome execution costs. Hence, a larger hedging interval is required. So my query was pertaining to hedging frequency not being irrelevant as I took your post to imply. Either way, I ain't makin' the rent with this strategy MoMoney.
Momo -- replicating Merton's "continuous time" BS is an academic absurdity and isn't practical. I understand you point, but "replicating" in "chunks" doesn't violate most pricing assumptions.
I thought one should use "odds to reach" and not only "odds to close" for this strategy , no ? So I build a simple binomial model where I am keep on changing the leight of the tree's branches to find the best intraday point to adjust. Most of the analytical software ignoring the intraday's hi-lo when calculating HV(SV). Man , looks like I am on the wrong track here.
Replication error is linear to expectancy if realized statvols = implied vols, disregarding spreads and commission. An decrease in "replication error" is + when statvols < implieds. Therein lies the "art" to trading gamma. It's a struggle between freq and size.