TY - THES A1 - Hülkenberg, Simon Cornelius T1 - Monte Carlo Tree Search for imperfect information games with the example card game ’Gwent’ N2 - Monte Carlo Tree Search (MCTS) is a common approach for the solution of imperfect information games. This thesis assesses the effectiveness of MCTS in the imperfect information card game Gwent. The assessment is performed with an Artificial Intelligence (AI) Gwent agent implementation based on the MCTS algorithm. The general effectiveness of the implemented AI agent is confirmed by playing against the official Gwent PC minigame. To assess the MCTS effectiveness, simulated Gwent games between two AI agents were conducted. Possible enhancements of the MCTS algorithm were tested in simulated Gwent games between an AI agent using the base version of MCTS and an AI agent using enhanced MCTS versions. Important parameters that influence MCTS were optimized empirically based on the simulation results. KW - Monte-Carlo-Simulation Y1 - 2020 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:bvb:860-opus4-2334 ER -