It's one man's opinion, but illuminiating nonetheless. I'm talking about Nate Silver's reviews of the eight daily tracking polls out there (i.e., their strengths and weaknesses).
For example, Zogby's daily tracking polls assume that the voter turnout for Democrats and Republicans will be the same as it was in 2004 (where Dems and Reps voted in roughly equal numbers). But is that a safe assumption? This year, there appears to be a five to ten percent increase for Democrats in terms of party identification, compared to 2004. Therefore, Zogby's poll results could be underestimating Obama's lead.
It's a good read, but for those who want the bottom line, Silver thinks that Rassmussen's daily tracking poll is probably has the best (or, better said, the least flawed) methodlogy.