Hull has a comment somewhere in his book saying something about the inadequacy ofExcel's Solver for some optimization applications.Does anyone know exactly what algorithm is used by Excel's Solver?

not sure but it does seem to be some type of closed form itterative process.i have noticed that i cant use it in binomial if trying to gola seek on certain barrier options...annoying at times.

I think it is generalized reduced gradient (specifically GRG2). But, it depends on whether you have "asssume linear model" checked in which case it will be a linear simplex.and then with binary/integer variables maybe some sort of branch/bound type model.

Have a look at:http://www.utexas.edu/courses/lasdon/design3.htmBM

mohamedb -- thanks a lot for the reference. It is always interestingto look at the designer's account of their design.The main point to take home is that the Solver's optimization algorithm, mentionedin this thread by mrowell as GRG2, finds local extrema, basically following the directionof some gradient until the direction becomes effectively zero (something in this spirit).In the fourth edition of Hull, p. 376, he comments as follows:"A general purpose algorithm such as Solver in Microsoft's Excel is liable to providea local rather than a global maximum....a special purpose algorithm, such as Levenberg-Marquardtshould ideally be used."The comment is made in the context of the MLE method, where a global maximumfor the likelihood function is sought. To me, this seems to be very good reason whynot to use Excel's Solver.

anybody have a resource that outlines the basics behind optimizing the MLE???

I've talked with the people who wrote the solver for Excel. They tell me that MS won't let them fix known issues.