Does anyone know how IB's options model works? Does it use a Heston/SBR model or BS and local vol using sticky delta? They wouldn't tell me....

I can tell you that IB Risk Navigator has major differences with OptionVue in calculating the Greeks. IB shows my July M3 having D/G/T/V as -2/-3/166/-521 while in OV they are -18.34/-2.38/224/-539.9.

ACS thanks for the feedback. I use a BS sticky delta model and my results never agree 100% with IB's. Good to know I am not alone. Do you know which model OV uses to calculate the Greeks?

Hi Hugh, Here's a screen shot of the OV model settings: The Yates model is proprietary but I think what model OV users use.

Thanks Tom, I'll take a look at OV. My understanding is that the Greeks calculated at T0 (same spot, option price, IV, DTE) should be the same for all models. But calculating the Greeks for subsequent days and different spot prices is where they will differ. Am I right?

ACS, Tom how does OV's T0 PNL curve compare to IB risk navigators? My model is similar to IB's but not exactly the same.

Hugh, could you please explain what you mean with "sticky delta"? Isn't IV the major contributing factor? I would like to implement an option model, but I'm not a quant. Any pointers (also from others) are highly appreciated. I mainly trade complex positions. Thanks.

yes but how do you calculate IV for different spots? constant vol isn't what happens in reality. Take a look at Derman's lectures http://www.emanuelderman.com/media/euronext-volatility_smile.pdf and his other lectures on his website. Once you have IV you can calculate the rest of the greeks

i have written an excel/vba model. Pros are that I can change the model as i learn more, cons are that it's really hard work. This is a good book http://www.amazon.com/Complete-Guide-Option-Pricing-Formulas/dp/0071389970

There will be differences in the current Greeks between software. TOS and ONE show different values from OV for example. That difference will vary from minor to major depending on market conditions. In my example above you can see that OV and IB show current Deltas that are far apart.

I agree the numbers are different, I just wasn't sure why. If all models use a BS formula the values should be close. So I guess they use different models like BS, Heston, SABR etc. Any ideas on what the large brokerage firms use?

Large brokerage firms use their own home-grown and proprietary volatility models (and risk models). My guess is that depending on the trading instrument and/or length of the trade, they could be using multiple types of diffusion volatility models. OptionVue uses CEV volatility model that uses stochastic volatility.

Agree. This is a really good book. It comes with a CD containing VBA codes for the different models. Very convenient for customization. Helped me a lot in understanding the pros and cons as well as the limitations of modeling.

Have you written your own model? If so how are you modelling the payoff chart/calculating options prices for subsequent days and spot prices for hedging. I use sticky delta and a BS model but have numerous questions. Once this model is developed a bit further I'll also use SABR as another curve.

No I haven't done my own model. But I modified the VB code provided by the book to derive some greeks that are important but not available on TOS - eg. dDelta-dTime (Charm) and dDelta-dVol. I combine them with the standard greeks to assist my decision making but I haven't done anything to try to merge all the factors into a single indicator. The problem with any one mathematical model, in my opinion, is that in the real market, there are factors that are difficult to model accurately. For example - few years ago, I read this article by Mark Sebastian on stages of the IV skew http://www.optionpit.com/blog/stages-spx-skew In my experience, this turns out to be a highly critical factor that affects option pricing and more specifically for us the accuracy of the T+0 line. Frequently, I noticed the T+0 line changes its shape quite drastically. What I thought was a safe position yesterday isn't so anymore today. I realized that has to do with the changes in the shape and the skew (steepness) of the IV curve. If my understanding is correct, although OV incorporated the volatility curve and CEV into their calculation, but their model does not take into consideration the changes in the IV skew, ie they assume the steepness stays constant. In real life, the skew moves around quite a lot. Not simply just the steepness but also the shape of the curve changes as well, as shown by article. Therefore, I spent some time gathering empirical data on RUT skew to understand the behavior and to use this knowledge to improve the accuracy of my adjustments. It does end up changing some of my decisions from time to time.

Kevin you are 100% correct. I'm starting to get my head around modelling option pricing. There are either jump diffusion models (Heston or SABR) or stochastic local volatility models (SLV) that we are discussing here. I like the SLV approach. You are right, if the IV surface doesn't change then hedging would be easy - T+n prices would be easy to calculate based on an options moneyness (K/S) and its IV (a plot of moneyness and IV is sort of stable). So far I know you can either directly predict/guess changes to the IV surface or convert it into a local vol surface using Dupire's eqn. Emanuel Derman has also done extensive work on the vol smile (check out his website). As a stop gap I've increased the IV surface curvature (more conservative prices, higher delta) until I have a better predictor model or the local vol surface gives a better result. I've asked a couple of questions on quant exchange, but no feedback yet. Is this something you are interested in?