I'm trying to improve my methods for calculating real-time US Equity option portfolio risk.
My main problem is volatility "stability" across all strikes in an option series.
The current implementation is similar to the OCC Haircut calculations, where I calculate volatility at current market prices and then calculate theoretical prices for all options at 10 equidistant percentage moves of the underlying up and down.
My problem is generally with deep ITM/OTM options that have a short time to maturity and/or wide bid-ask spreads, no bids, etc. I often see call vs. put IV's severely out of whack which in turn causes deltas to be incorrect (especially for portfolios that do large reversals/conversions for dividend plays). I'm using a dividend forecast feed in my current pricing models (Escrow method).
I was wondering if it would be better to extrapolate the skew curve from the ATM options to estimate volatility at further strikes? Are there any programmatic examples out there that describe and handle this situation?