Idea behind default minimization strategy

[acampove]

Hi,

I am going through this part of the documentation and It is not clear why we would need the calculation of the Hessian in the minimization step, i.e. strategy 1, which is the default. Shouldn't Strategy 0 be all we should need? The Hessian should only be needed at the end of the minimization, to get the errors. I am saying this because I see a significant improvement of the speed in the fits with strategy 0.

For more details check this issue. I tried to reproduce the same with iminuit but I am not an expert on that yet.