There are two types of ball interception, referred to as active and passive interception. The passive interception is used only by the goaltender in some particular cases, while the rest of the team uses only the active interception. Each cycle, the interception target is recomputed so that the most up to date information about the world is used.
The active interception is similar to the one used by the Humboldt '97 team[3]. The active interception predicts the ball's position on successive cycles, and then tries to predict whether the player will be able to make it to that spot before the ball does, taking into account stamina and the direction that the player is facing. The agent aims for the earliest such spot.
This process can be used for teammates as well as for the agent itself. Thus, the agent can determine which player should go for the ball, and whether it can get there before the opponents do.
The passive interception is much more geometric. The agent determines the closest point along the ball's current trajectory that is within the field. By prediction based on the ball's velocity, the agent decides whether it can make it to that point before the ball. If so, then the agent runs towards that point.