This paper demonstrates the use of recursive modelling of opponent agents in an adversarial environment. In many adversarial environments, agents need to model their opponents and other environmental objects to predict their actions in order to outperform them. In this work, we use Deterministic Finite Automata (DFA) for modelling agents. We also assume that all the actions performed by agents are regular. Every agent assumes that other agents use the same model as its own but without recursion. The objective of this work is to investigate if recursive modelling allows an agent to outperform its opponents that are using similar models.