RE: directionality input in option pricing I am assuming that a lot of otions marketmaking is automated with some presumably sophisticated option pricing models. I wonder if the models include some directionality besides purely statistical/volatility inputs. A human trader presumably always includes directionality aspect in his calculation, so should not there be an edge for him against a purely statistical model?
Automated market making models should be sophisticated. They should be able to have directional biases too. I'm assuming you are talking about a bias in volatility and not a bias in the movement of the underlying.
I actually meant bias in direction/trend of the underlying. But I guess bias in vol trend should also be part of the formula to be competitive. So you are saying all automated market making algos do take all that into account and reprice all offers constantly?
market makers (automated or non automated) do not have a bias in the direction of the underlying. If they did have a view on the underlying, then they can trade the underlying without options.
The MM has a more hollistic/multi dimensional view than that. They actually try to predict the evolution of the whole volsurface. Thats skew x forward curve. Once they plug in their pricing curves, its just a matter of setting quantity limits so that they don't pick too many deltas all at one time. Some systems can even setup levels to back off the market. but keep quoting. That way, they can still harvest fat finger limit orders or market orders.
I would think the inventory level of their book would dictate a delta leaning, but not trending tools.
Stochastic vol models have a correlation coefficient describing the relationship between the Brownian processes driving the vol/spot variables. Take the SPX for example. VIX usually increases when the market declines. Most would model this as a negative vol/spot correlation, meaning that a short ATM straddle is actually inherently long biased. It may appear delta neutral when priced via Black-Scholes, but since you are pricing in the presence of a smile, it's truly not a flat position. In the real world you can never remove directional risk entirely. Drift matters if you're trading options, even delta neutral. To get "purified" vol/var exposure, you need a swap. In terms of the edge you speak of, no I doubt it. Even us guys using old fashioned models like Black-Scholes are aware of this. It's just easier to mess with the one vol parameter to adjust prices/hedge ratios. Best
so are you saying the short atm straddle would be considered inherently long because of the effect of vega?
Yes, but I wouldn't call it that. And I'll emphasize, by no means am I saying a short straddle won't take losses if the underlying explodes higher. Just that the negative correlation between spot and vol for indexes means that short index vol is somewhat long biased. Vega is only a measurement. It doesn't cause anything. So I would say "because of (index) spot's negative correlation with stochastic (index) vol."