I mainly observed this empirically where my basic BS calculations wouldn’t match what happens during large market swings, without understanding of why or what specifically fails, but simply deriving my own Greek-free correlations and models. While some people like Ron Bertino and the guy running SJOptions do seem to calculate how Vega changes differently across calendar expiry dates with underlying’s movement, and I’ve seen some mentions of “True Vega” or maybe “Dynamic Vega” with reference to Nassim Taleb’s book “Dynamic Hedging”, so possibly that book contains more info (I don’t have it).
Generally BS doesn’t even predict volatility smile, so it’s been adapted/utilized to more realistic market conditions.
While what I have observed is that, for example, as IV swings wildly and then settles at a high level, some of my ratio spreads keep changing value for a brief time even while the IV already settles and stops changing too much. This cannot be explained by any greeks or calculations.
I even had DOTM ratio spreads offset by equivalent back ratio spreads set couple weeks apart (both about 3 months out), and at one point they were both losing money as if both using completely different models or calculations. Though the following day they’ve equalized to the point of making sense again.
Some people would interpret this as market makers conspiring to kick them out their positions. While I was actually wondering whether MM’s options inventory doesn’t play a role here. I’ve also noticed very small differences between pricing of weekly vs monthly DOTM options, as if they’re handled by different MMs, but I’m not sure.
Though AFAIK for those reasons some quants seem to study vol surfaces empirically rather than trying to calculate them using BS or other static models (?).
While I’ve sometimes spent 15+ hours a day staring at options chains and I’d like to think I’ve derived some conclusions that I can use, while still suspecting that MMs may have more control than some people may think, and can for example set opening DOTM options prices after a market crash at whatever price levels they need and how it may affect their own positions/inventory (thus producing or skewing the vol smile), not at levels that can be pre-calculated. This may produce various inefficiencies as well as risks. Though I may be dumb and wrong, so I write software that assumes that I am dumb and derives its own correlations and conclusions empirically, without my personal opinions/biases, and I get much better (actually amazing) results when I remove Greeks as variables.
