Coarse grained dynamical entropies characterize in a quantitative way the p
recision-dependent predictability of a system. This gives a much more compl
ete description than solely the Kolmogorov-Sinai entropy. The concept is ap
plied to two classes of high-entropic systems, to one class which simulates
stochastic processes in some suitable limit, and another class of extensiv
ely chaotic spatially extended systems. In both cases, the coarse grained e
ntropies yield new insight into the length-scale-dependent complexity of th
e systems. (C) 2000 Elsevier Science B.V. All rights reserved.