The race was even among likelies in CNN’s poll after the GOP convention, so here’s further confirmation that O has indeed gotten a bounce. (Of particular note, his favorable rating rose to 57 percent, his highest in two years.) Three major caveats, though. One: While there’s a nice bounce among likely voters, there’s virtually no bounce among registereds. Last week O led Romney by seven points among that group; this week he leads by eight. Obviously the composition of CNN’s sample of likely voters has changed, although because they never publish their partisan splits, there’s no way to tell by how much exactly.

Speaking of which, number two: Romney trails overall by six points while leading among independents … by 14?

The only way to make those numbers make sense is if the sample of likely voters skews sharply Democratic, which seems improbable, and if the sample of independents here is minuscule. That wouldn’t be unusual for CNN, if so: Just like last week, it looks like they’re pressing hard to get self-identified indies to identify themselves as leaners one way or the other and then reserving the indie column for the few remaining “true independents.” With a small enough sample, Romney could theoretically lead O by 50 points among this group and still trail overall.

Three: There sure does seem to be a pronounced age split in this poll:

That’s right in line with conventional wisdom — O leads big among younger voters, Romney leads among older — but note that it’s not young adult who are really driving Obama’s advantage here. It’s the middle-aged, ages 35 to 49, that are decisive to his overall margin. Not sure how to explain that dramatic effect. It’s always tempting to resort to Medicare politics to account for sharp age disparities but it’s not obvious to me why near-seniors would be fairly evenly split while slightly younger voters would tilt sharply. Maybe the Democrats’ culture-war convention played well with them? But middle-aged adults tend to be pro-life on balance. Hmmm.

Lots of oddness to O’s bounciness, then — although, bearing that in mind, I still don’t like the look of this trend, especially in light of the last post:

This one strikes me as bad too, although there’s nothing surprising about it:

Romney’s been banking since day one on winning an “Obama vs. Not Obama” election, but it’s easier to imagine a race like that ending in a semi-landslide for the challenger than in a narrow victory, which increasingly appears to be Mitt’s best-case scenario. The “Not Obama” voters in Mitt’s column will be out in force no matter what, whereas O’s support among pro-Obama voters depends, I think, to some extent on how the polls look down the stretch. If it’s close, he has a stronger pitch to “unlikely voters” who like him to haul their asses down to the polling place on election day and put him over the top then he does if the race looks like a lost cause circa mid-October. But then, this is really just another way of saying that Romney needs to do more to make the pro-Romney case among undecideds rather than just the anti-Obama case, which is already as clear cut as can be. That’s what the convention was supposed to be about, and per the trendlines here, evidently it did some good. Enough good, though?