[romanows]

They're clearly taking a reinforcement learning approach. Just a few thousand more returns and you'll be good to go!