Kos posts some "expert" (scare quotes his, though I condone them here) analysis that reches some encouraging conclusions. Basically, Bush would get trounced if the election was held today and something big needs to happen for Bush to pull this one out. you never know though...

If you're more comfortable basing your judgement of the state of the race on statistics, things still look good for Kerry based on this analysis:

Counting the last three polls the current probability of a Kerry win is 98.6 percent. 270 electoral votes (EV) are needed to win. The median (50th percentile) outcome is Kerry 307 EV, Bush 231 EV. The 95 percent confidence band for Kerry is 275-333 EV.

Current probabilities by state (rounded to the nearest percentage point): AR 3, AZ 11, FL 87, IA 100, ME 90, MI 100, MN 100, MO 47, NV 53, NH 100, NM 100, OH 7, OR 100, PA 97, TN 69, WA 100, WV 64, WI 97.
The rank order is currently (boldface indicates states statistically in play at the 20-80% level):
D <- IA/MI/MN/NH/NM/OR/PA/WA/WI (95-100%) - ME - FL - TN - WV - NV - MO -AZ - AR/OH (0-5%) -> R

And since it's always fun to add graphs to the site, here's a particularly pretty one:

But is Ohio really 93% likely to go for Bush? Is Tennessee really 69% for Kerry? Neither sounds too likely (in addition to some of his other numbers). In light of this, my own preference is to combine the two and use a statistical probability analysis but rely on my own views rather than polls (although polls would contribute to my views of course) to determine the likelihood each individual going one way or the other. Unfortunately, although I'm quite good at coming up with biased opinions of voting trends, I wouldn't know how to go about figuring out the probablity of national results based on state probablities. Fortunately, this site helps out. It starts off with his own predictions:

His: Bush 248-Kerry 290, 81% prob. Kerry wins

I did make a quite a few changes, but it must have pretty much evened out in the end:

Mine: Bush 247-Kerry 291, 81% prob. Kerry wins

I also put in the poll guy's probabilities, rounding to the nearest 10%, unless a number need to be rounded to (and was not at) 0% or 100%, in which case it would be rounded to 10% or 90% to leave some chance of an upset, and I didn't round to 50%, but instead to 40% or 60% in order to show that there was a slight favorite in those states. Poll guy didn't analyze CO, LA, or VA, so I assume he's taking them as 0% (i.e. 100% Bush). I'm not sure if the two are using the same methods but the results on this page should be pretty close to his right? Yep:

Kerry 306-Bush 232, 98% prob. Kerry wins

I hope the polls are more accurate than my biases (based on that phrasing it sounds like they should be, doesn't it?), but either way, it looks good for Kerry.