<p>Coalitions play a central role in influencer marketing, enabling businesses and government to leverage networks of authentic voices to disseminate targeted messages effectively. In this paper, we develop a mathematical framework for identifying minimal winning coalitions of influencers that ensure maximum follower engagement at minimal cost. By introducing predetermined relational dependencies and independencies among agents, we convert traditional simple games into relational games, employing combinatorial tools such as closure operators and choice functions. This approach yields several polynomial‑time algorithms for computing minimal winning and maximal losing coalitions. Building on this foundation, we extend the model to incorporate uncertainty through Bayesian relational games with probabilistic decision tables grounded in Pawlak’s rough set theory. We propose a repeated two stage game in which influencers first form coalitions ex-ante, selecting partners based on probabilistic assessments of relational closures, and then decide ex-post whether to remain in or deviate from existing coalitions. Observations of coalition outcomes inform agent’s type beliefs and update their decision tables in a reinforcement‑learning framework. For risk‑neutral and risk‑averse agents, we demonstrate that checking core emptiness at the ex‑post stage reduces to a linear‑feasibility problem, offering scalable methods for stability analysis under uncertainty.</p>