We review the equivalence of maximal center gauge fixing to the problem of
finding the best fit. to a given lattice gauge field, by a thin vortex conf
iguration. This fit is necessarily worst at the location of P plaquettes. W
e then compare the fits achieved in Gribov copies generated by (i) over-rel
axation, (ii) over-relaxation after Landau gauge preconditioning, and (iii)
simulated annealing. Simulated annealing yields the best fit if all links
on the lattice are included but the situation changes if we consider only t
he lattice volume exterior to P plaquettes. In this exterior region, the fi
t is best for Gribov copies generated by over-relaxation. and worst for Gri
bov copies generated after Landau gauge preconditioning. The two fitting cr
iteria (including or not including the P plaquettes) yield string tensions
differing by -34% to +20%, respectively, relative to the full string tensio
n. Our usual procedure ("quenched minimization") seems to be a compromise b
etween these criteria, and yields string tensions at an intermediate value
close to the full string tension.