Endgame Solving in Large Imperfect-Information Games