The Design of Spoils: AI and Gameplay

The AI in Spoils was designed to be a friendly and non-threatening opponent that could make mistakes, creating a casual and enjoyable experience for players. The AI system used deep reinforcement learning to train itself on the game, receiving rewards for winning tricks and playing good moves. Both opponent and partner AI were trained using the same system and received rewards for individual actions as well as teamwork. The bidding phase was more heuristic-based, as quantifying the goodness or badness of a chosen sweeper was challenging. The AI training process allowed for the discovery of interesting strategies and provided feedback on the design of the game.