[ ] stats means can be produced by applying the weight proportional to the move/line relative playing probability (move_freq)
Each repertoire can be classified with several numbers:
the yaml settings of course
a weighted mean of the "leaf" of the repertoire that is the sum of the scores of the last moves of each lines weighted for the number of games in that position divided by the sum of weights
it can be done both for the db score (black or with depending on the repertoire side) and for the engine evaluations
Produce a report at the end of the repertoire generation. The report should include the plots and the stats of the moves.