How AP uses expected vote instead of 'precincts reporting' when determining a winner