We discuss various experiments on the time decay of velocity autocorre
lation functions in billiards. We perform new experiments and find res
ults which are compatible with an exponential mixing hypothesis first
put forward by Friedman and Martin (FM): they do not seem compatible w
ith the stretched exponentials believed, in spite of FM and more recen
tly of Chernov, to describe the mixing. The analysis leads to several
byproducts: we obtain information about the normal diffusive nature of
the motion and we consider the probability distribution of the number
of collisions in time t(m) (as t(m) --> infinity finding a strong dep
endence on some geometric characteristics of the locus of the billiard
obstacles.