The paper studies strategies represented as finite and probabilistic automata in infinitely repeated «Prisoner’s Dilemma». A comparative analysis of the strategies is provided. It includes a comparison of players’ payoffs obtained when playing against a particular automaton or against a randomly chosen automaton. The terms of frequency of cooperation, semi-cooperation, and non-cooperation are also introduced and calculated for each pair of strategies under consideration. A subgame perfect equilibrium conditions for some strategy profiles, not previously studied in the literature, are obtained. All theoretical results are illustrated by numerical examples.