SimSoccer Coach: Soccer coach simulation

• Main Authors: Dulalia, Conirose L., Go, Peggy Sharon L., Tan, Pamela Vianne C., Uy, Maria Zaide Ilene O.
• Format: text
• Language: English
• Published: Animo Repository 2004
• Subjects: Soccer > Coaching > Computer simulation; Soccer > Training; Soccer > Rules; Soccer players; Computer Sciences
• Online Access: https://animorepository.dlsu.edu.ph/etd_bachelors/14254
• Institution: De La Salle University

The SimSoccer Coach system shows single agent learning by analyzing the fixed opponent's behavior and providing offensive and defensive advice to improve the team's performance. For the offensive advice, the system imitates the successful passing and shooting behavior of the fixed opponent's previous opponent. For defensive, the system observes the opponent's passing behavior and counters providing advice for marking players and intercepting the ball. To generate these sets of advice, the system reads logfiles of previous games played by a fixed opponent against other teams and selects the data to be used for learning. C4.5 decision tree algorithm is used to construct the tree and generate production rules based on the selected data. These production rules are converted into Clang advice following the Coach Language grammar. These Clang rules are then given to the coachable team before the game. After the game, the system outputs game statistics and list of rules fired during the game. Tests using different rule accuracy thresholds proved that the team's overall performance improved through the offensive and defensive advice given. Further comparison against RoboCup 2003 coach entries showed that the SimSoccer Coach System can compete with other coach entries.