Version 0.6.6
Kristaps DzonsonsFix the simulation mode for best-reply, which had a bug in computing payoffs. Also, add a Peyton learning model (Young, 1998) option to the best-reply, which modifies the best-reply function to, instead of accepting the current population shares per players' strategies, accept a frequency of past play by drawing from a window of past plays. Also clean up the mathematics of the Operator's Manual to greater consistency of notation.