I'm calibrating heston with the S&P500 options. Can someone provide the initial guess of the parameters that I can start with? Thanks.Wondering

rho = -0.7vol. of vol. = 0.5intial vol. = long run vol = VIX^2mean reversion = pick a small integerregards,

Thank you, Alan.One dumb question, since S&P500 pays dividend, how should I revise the formula of the vanilla call? I thought that I should use (r-q) to replace all r in the formula of P(1) and P(2). And the Call option should be S(0)*exp(-q)*P(1) - K*exp-(r-q)*P(2) instead of original formula ofS(0)*P(1)-K*exp(-r)*P(2)

Not quite, but if you replace S(0) everywhere by S(0)e^(-q T), should work

Thanks a lot, Alan. I got itBTW, if VIX^2 is a good place to start, does it mean that both inital vol and long run vol are not quite stable over time? Does your book on Stoch Vol covers Heston and related implementation? Thanks.Wondering

QuoteOriginally posted by: Alanrho = -0.7vol. of vol. = 0.5intial vol. = long run vol = VIX^2mean reversion = pick a small integerregards,Which way do you use to choose your initial estimate? MC simulation?

QuoteOriginally posted by: wonderingThanks a lot, Alan. I got itBTW, if VIX^2 is a good place to start, does it mean that both inital vol and long run vol are not quite stable over time? Does your book on Stoch Vol covers Heston and related implementation? Thanks.wondering,The initial vol is not supposed to be stable; it's like the stock price. If the model were correct,the long run vol would be stable but, you're right, it's not. Regarding the book,it does treat the model, but I'm also treating it in more depth in Vol II (forthcoming) covering things likethe branch cut crossing problem. p.s. if you don't mind, why don't you post your results? (and the min/max expirations you included). CB, where do the guesses come from? -- you just run a few and look at other people's results.After enough of those, you have an idea.

Thanks again for your guidance, Alan. I’m having a hard time with calibration. I have vanilla call prices for maturities from 1 month to 10 yr and strikes ranging from 50%-200%. I have the VBA code of Heston in Excel and I basically used Excel built in Solver to calibrate. There is a paper by Nogel on Heston model basically saying that if we get good the initial guess, the Excel solver actually works pretty well. I have yet been able to get some reasonable results by using Excel solver. Do you think I should fit the model with shorter maturity and more narrow strike range? The option I want to price do have very long maturity.

I am not an Excel user, so can't help with that.Are you fitting to prices or implied vols? (I always fit to implieds). Are you using bid/ask midpoints? (I would)Are you throwing out all no-bids? (I would)You might want to start with just two maturities, say around 3 months and 1 year. Are your implieds reasonably smooth vs. strikes?

I got implied vol quotes from a trading desk which I think are in good shape. I also built out my own implied vol surfaces by backing out the vanilla prices traded on CBOE(within 3 yr maturity). These two are pretty consistent. I will take your suggestion of using mid prices and throwing out strikes w/o any bids.I used 2 maturities as you suggested. The results look reasonable and fit pretty well. However, I don’t have much data points. Here is what I got anyway.V_0 0.0128 Theta 0.0375 Kappa 0.9933 Sigma 0.3167 Rho -0.7178 I then switch to the method of calibrating implied vol instead of prices. My problem is that in some iteration, the implied vol failed (weired prices). Then, optimization stops because the sum of squared error calculation on implied vol fails. How should I deal with this? Should I set the squared error to a really big number for that cell whenever implied vol failed for that cell?Thanks a lot for your help.Wondering

Play around with it for a couple of days, trying things like you just suggested.If you can't get it to work, report back what you have tried.Some things to focus on: patently illegal vs. legal parameter valuespatently wrong vs. legal returns from the Heston model pricerpatently illegal vs. hard-to-handle input to the implied vol. routine

alan,does the particular set of initial parameters you suggested herefor the heston model depend in some way from the option wondering was going to price or is it reasonable for every kind of option?And do you think a local calibration ,i.e. a calibration starting from a guess of the parameters,is preferrable to a global one?What about the risk of ending in a local minimum without even knowing it??I 'm deciding how to perform the calibrationsky

QuoteOriginally posted by: skyrmionalan,does the particular set of initial parameters you suggested herefor the heston model depend in some way from the option wondering was going to price or is it reasonable for every kind of option? .......... It was based upon the SPX, but probably would work for most if the optimizer was decent. And do you think a local calibration ,i.e. a calibration starting from a guess of the parameters,is preferrable to a global one?What about the risk of ending in a local minimum without even knowing it?? ......... You should always re-start the (local) optimizer at different values. If it keeps converging to the(more-or-less) same results, I would have confidence. If you can get a true global optimizer to work, great, although that may be wishful thinking.

i write a vba routine to use excel solver but with the function solverok i must choose the target cell and value, the changing cells and the min/max constrain (how can i use that?)but i don't know how i can write (in vba) the constrain on the parameters...can anyone help me?i'm writing the calibration code for some levy model and now i will do it for heston: a question how much time your code use to get the parameters?my is 30-60 seconds but depends by initial valuesbye

Should have a look at the page : http://support.microsoft.com/kb/153442/en-us