>>11469682Well, using only historical data will have the problem of black swan events, although maybe you can simulate them.
Student's t-distribution seems good. Is there any way to determine how many degrees of freedom, or should I just attempt to fit all the coefficients? Some studies I found fit it to n = 2.5 or so, which does not seem all that analytically robust.
I'm using it for options pricing.
>>11471364That doesn't have fat tails.
