Maximizing votes under imperfect information