乒乓游戏的奖励——（OpenAI 健身房）

Question

I know that the Pong Game initializes to new game when one side scores 20 points.

By the way, the reward shows that it goes down below -20.
Why is that so?

One thing to expect is that after one side gets 20 points, the game is reset by playing one more time. Does the game need to get 21 points to initialize?

(Use 8 workers, A2C, PongNoFrameskip-v4)

Answer 1

Pong is played to 21 points, not 20.