Manipulation of journal metrics such as the Impact Factor (IF) through editorial decisions that increase self-citations has been empirically observed in recent studies, sparking discussions on how to prevent such practices. In this work, we present an agent-based model where journals are rational agents seeking to improve their positions in the journal ranking. This allows for the exploration of these issues by computational simulation. The model generates synthetic citation networks that reproduce key patterns of academic social networks, such as the Matthew Effect, and accurately reflects secondary emergent patterns like the specialization of manipulation strategies across disciplines. We compare two IF calculation policies—one including and one excluding self-citations—and show, through statistical analysis, that removing self-citations reduces the incentives for manipulation. These results provide tools for developing policies that promote scientific integrity.