Hi Benny,

check this code line: returnG = list(map(lambda s: s.reward * (GAMMA ** len(s.steps)), elite_candidates))

The value of total return depend of the episode length

regards,

Jordi