Here's Mark Blumenthal's summary:

Here are the daily tracking results from today. All changes are 1-2 percentage points on the margin, which is well within the usual range of sampling noise. If we use the "expanded" likely voter model for Gallup (our usual practice), we have four tracking moving slightly in McCain's direction, two in Obama's direction and two with unchanged margins. If you prefer to watch the "traditional" Gallup likely voter model, the count is 4 to 3. Once again, this looks mostly like random variation.

We want to hear what you think about this article. Submit a letter to the editor or write to letters@theatlantic.com.