Reverse Engineering and Experimental Analysis of a Deterministic Enemy AI in Pac-Man

Early arcade video games relied on deterministic, rule-based artificial intelligence due to extreme hardware and memory constraints. Despite their simplicity, these systems produced complex and engaging enemy behaviors that continue to influence modern game AI design. This paper presents a reverse engineering and experimental analysis of Blinky, the red ghost in Pac-Man. Using a modified open-source implementation with step-controlled execution, Blinky's movement behavior was isolated and analyzed at each decision point. Experimental results show that Blinky follows a greedy Manhattan-distanceminimization strategy during its chase mode, while operating under a distinct roaming behavior prior to chase activation. The study demonstrates how simple deterministic rules, behavioral modes, and maze constraints interact to produce emergent pursuit behavior.