OH, so they are accrate in presedential elections, but using the same methodology have a 30% error rate in other elections Makes sense to me