Deep reinforcement learning for optimizing penalty kick strategies in football: a comparative study of PPO and IPPO

Ahmad Naim, Muhammad Fadhlul Wafi and Mohd Ibrahim, Azhar and Zaid, Mohd and Ahmed Allam, Ali (2026) Deep reinforcement learning for optimizing penalty kick strategies in football: a comparative study of PPO and IPPO. In: 2026 IEEE 5th International Conference on Computing and Machine Intelligence (ICMI), 8-10 April 2026, Saudi Arabia.

	PDF - Published Version Restricted to Repository staff only Download (1MB) | Request a copy
Preview	PDF - Supplemental Material Download (189kB) | Preview

Abstract

Penalty kicks are among the most decisive and pressure-filled moments in football, often shaping the outcome of matches. Their complexity lies in the split-second decisions made by both kicker and goalkeeper, which traditional statistical or biomechanical approaches struggle to fully capture. In this paper, we explore how Deep Reinforcement Learning (DRL) can optimize penalty kick strategies using the Google Research Football (GRF) environment. We trained agents with two algorithms, Proximal Policy Optimization (PPO) and Independent PPO (IPPO), to model realistic kickergoalkeeper interactions. Performance was analyzed through goal success rates, goalkeeper saves, and decision heatmaps. The trained IPPO agents exhibit a goal success rate exceeding 85%, a notable increase compared to baseline strategies which uses PPO agents (~65%). Spatial heatmaps further revealed a strong preference for low-corner shots, aligning with real-world tendencies. These findings show that DRL-trained agents can learn adaptive and effective penalty strategies, offering practical insights for football analytics, training, and tactical preparation.

Item Type:	Proceeding Paper (Other)
Uncontrolled Keywords:	GRF, PPO, Decision-Making, Football Simulation
Subjects:	T Technology > T Technology (General)
Kulliyyahs/Centres/Divisions/Institutes (Can select more than one option. Press CONTROL button):	Kulliyyah of Engineering > Department of Mechatronics Engineering Kulliyyah of Engineering
Depositing User:	Dr Azhar Mohd Ibrahim
Date Deposited:	21 Jul 2026 11:46
Last Update:	21 Jul 2026 11:46
Queue Number:	2026-07-Q4037
URI:	http://irep.iium.edu.my/id/eprint/129902

Actions (login required)

View Item