For now, use most current rankings. Preseason polls based on projections shouldn't matter. Early season polls based on limited body of work shouldn't matter. The most recent polls after people have seen more and more of these teams is what people should look at as far as "good" wins. Do you consider Oklahoma losing to an unranked Wisconsin team as a bad loss, but Rutgers losing to them isn't since they were in the top 25 by that point? They played the same team (obviously injuries can make a pretty big difference but I'm mainly talking about teams being overrated earlier in the year).