Although, if I were going to cherrypick a poll, I'd pick one that proved accurate in the last election (i.e., neither Rasmussen nor Gallup, which after all the votes were counted both ended up being on the lower end of the polling ranks).

Tying for most accurate were Democracy Corps, Foxnews/Opinion Dynamic, CNN/Opinion Research, and Ipsos/McClatchy. Of those, only Ipsos has polled since the first debate, and their latest poll (before last night's debate) shows Obama back in the lead and with the momentum.

Of course, even unbiased cherrypicking is still cherrypicking. It's still better to average all the recent polls. That does show Romney with the tiniest of leads, which is more likely to be random error than reality. And even that fails to account for last night's debate, which could go either way... but for which tracking polls suggest Obama won by a very small margin. So probably we are still at a statistical tie, though maybe with Obama having the illusory lead this time.

But then you've got to account for the Democrats' early voting advantage (much of which they got in while Obama's polls were soaring), for Obama's electoral advantage (which is slimmer now, but still present), and for turnout (which could do strange things this year).