usual tennis

config/trainer_config.yaml

     threaded: false
     time_horizon: 1000
     self_play:
-        window: 100
-        play_against_latest_model_ratio: 0.0
+        window: 10
+        play_against_latest_model_ratio: 0.5
         save_steps: 50000
         swap_steps: 2000
         team_change: 100000