An optimization engine identifies an infeasible node in a tree data structure that corresponds to a programming model, and computes a fake objective value for the infeasible node. The optimization engine then updates a branching variable pseudocost using the fake objective value. Next, the optimization engine uses multiple branching variable pseudocosts corresponding to multiple branching variable candidates in order to select one of the branching variable candidates. In turn, the optimization engine branches to the corresponding branch of the selected branching variable.