You are missing one important point here. Yes, vega is the sensitivity to changes in implied volatility, but the implied volatility doesn't rise or drop equally across different expiration months. That is, each expiration has it's own implied volatility or rather volatilities, so going into earnings or some other significant event, the front month options will experience the biggest volatility increase, aka volatility rush, and will also experience the biggest volatility drop, aka volatility crush, after the announcement. Volatility changes in back month options can be only a fraction of the front month's volatility change prior to and after earnings, so back month options, while having higher vega, will experience smaller price changes cause their implied volatility didn't rise and/or fall as much as the front month options' one. Here's an example, suppose current ATM implied volatility for the front month options on a certain stock is 30% and the back month is 32%. As we get closer to the earnings announcement the front month IV rises to 50%, while the back month IV to 35%. After earnings are announced the front month IV drops down to 25% while the back month IV falls to 29%. So as you can see, the front month IV went up by 2000 basis points and then dropped by 2500 basis points, while the back month rose only by 300 basis points and then dropped by only 600 basis points, so although back month options have much hihger vegas, the volatility changes were nowhere near those in the front month and hence the impact on the option prices was smaller.
To answer some of your replies: I know that IV calculation does not include Vega. BUT how do simulators like TOS calculate future IVs? You enter the percentage drop that you expect in IV and they model future IVs, and thus future P/L. How do they calculate this? My understanding is that they use Vega. Like they probably use delta to model a change in price.
MTE, you already said that, and I understand that different months have different IVs. In fact, read my thread about "BIDU earnings" and you will see that I explained how to model this in TOS. This is not the point here. I am going to ask one last time, and then I will just give up, as obviously there is a communication issue: "Can we trust Vega as a way to calculate future P/L around earnings, as it looks like options with higher vega do not automatically drop more than the ones with lower vegas, but this is what is used in calculators/modelers".
Simulators, aka models, do NOT calculate future IVs, they are simply theoretical price calculators, they take the inputs you give them and give you theoretical option prices based stock price, volatility, time to expiry and etc.
TOS gives you future (or "theoretical", are you playing with words, I am sure you know what I mean) IV's. How do you think they are "calculating" them?
Yes, they do drop by more provided their IV drops by the same amount as the IV of the options with lower vegas. In other words, given a fixed drop in IV, options with higher vega will drop more than options with lower vega, HOLDING EVERYTHING ELSE CONSTANT.
It's not the same, theoretical IV is the IV of the option based on the current market option price. TOS simply takes current market price and then uses Black-Scholes or some other model to find the implied volatility. When you change the volatility in the model you obviously change the option price.
I wouldnt rely on the way TOS models $vega PnL. To be accurate you need to model each strike and duration independently. TOS doesnt have that. They are a useless platform if you want to analyze vols.
In other words, given a fixed drop in IV, options with higher vega will drop more than options with lower vega, HOLDING EVERYTHING ELSE CONSTANT --------------------------------------------------------- Well, that was difficult, but I am glad you are finally agreeing with me on this one.
I never debated this issue. The point is that volatility rarely drops by an equal amount across expiries and strikes, also don't forget about a move along the volatility skew when modelling your PnL.