Livevol Pro

[kapw7]

Quote from sle:


(2) You can easily check if they are using business or calendar days for time calculation. Take any weekly option for next friday - the difference is year fraction is meaningful (4 business days vs 6 calendar days). The vendor that's showing higher volatility is the one using the business days, since he has a lower normalization factor (divisor). I preferr using trading day time, some people even adjust it to reflect the number of trading hours left, but that is usually an overkill.

This is something I find confusing. Is it not standard to use calendar days for IV calculations for example in the B-S formula, or the VIX calculation. Or even the T in the put-call parity or the forward calc formula etc? Do you actually mean that you use something like:

T=trading days/252 instead of T=calendar days/365

 

[sle]

Quote from kapw7:

This is something I find confusing. Is it not standard to use calendar days for IV calculations for example in the B-S formula, or the VIX calculation. Or even the T in the put-call parity or the forward calc formula etc? Do you actually mean that you use something like:

T=trading days/252 instead of T=calendar days/365

Yes, I use T = trading days / 252 for a variety of reasons, one of them is that variance settlement is based on trading days / 252.

 

[filthy]

in addition to what has already been said, the actually numbrical method used for the pricing engine can have large effects. for example, unless you are very careful, a tree will produce odd oscillations as it converges to the "true" result. and it is tricky to say whose model is best as you don't generally have the true result.

but these days most systems will be good enough. you won't get them to match up but the differences will be minor.

this isn't the case with large dividends though...

wrt the trading days vs calendar days argument for the time input into pricing models, the reason that it is still going is that neither side is right. option markets don't take time out smoothly like this. ie a weekend is neither an overnight move (as implied by trading days) nor a 2 day break (as implied by calendar days). it is somewhere in the middle. if you are trading one or two products you can adjust in your head. if you have a large book, the state of the art is to use a base time (252 or 365) to calc "dirty vol" then adjust each day by whether it is a weekend, vacation, slow day in mid-summer, earnings day etc to get a "clean vol".

i suspect sle knows all this. it is also totally impractical for anyone who isn't doing this full time.