This paper presents a parallel implementation of the adversarial search algorithms Minimax, Alpha-Beta Pruning and Expectimax in the Pacman game, based on the CS188 project from UC Berkeley. The contribution lies in transforming the sequential implementations into parallel versions that distribute subtree evaluations across multiple threads, reducing execution time. Experiments were conducted in a controlled text-mode environment to avoid external interference. The results show substantial reductions in execution time for all algorithms, with Minimax exhibiting the most significant gains. The study also compares the strategic effectiveness of sequential and parallel versions, highlighting how parallelism influences win rates. This work reinforces the practical value of parallel search in educational gaming environments and suggests that future research may combine parallelism with additional optimizations, explore deeper search trees, or adopt alternative parallel models.