This article investigates the dynamics of portfolio allocation within political parties to shed light on the patterns of conflict and cooperation between rival party factions. It provides a game-theoretic model that helps in explaining differences in portfolio allocation due to alternative modes of party organisation or party system competitiveness. Focusing on party congresses to estimate the number, strength and policy positions of party factions, the Italian case is analysed by testing some hypotheses generated by the theoretical model. The results shown that, overall, spoils are shared in proportion to the strength of each faction, in line with the prediction of Gamson's Law. However, there are also some important deviations from this path. Rules that foster party leaders' autonomy in fact provide them with a higher degree of discretion that will be used to reward their followers and to ward off any credible and harmful threat to party unity. Indeed, strategic portfolio allocation might balance out a lower amount of policy payoffs and becomes a strategy to restrain minorities from breaking away, thus contributing to the preservation of party unity in highly competitive political systems.