Here's a non-scientific analysis, but I believe a sound one, of what's going on in the Gallup 7-day horserace tracker, based on the movement in the 3-day approval tracker.
The bottom line: some really good Obama days have come off the Gallup 7-day tracker, while two really horrible ones have remained on there, with mediocre numbers since. The three-day approval numbers show what we know: there was a hit from the first debate loss, but that it's receding, and this should show up soon in the RV/LV numbers.
Here we go, after the jump.
We know that Oct. 7 and Oct. 8 were good days - Gallup broke out the polls for the first two post-debate days, and then the next two days, and its aid that RV was tied, then jumped to O +5 for those two days.
We see consistent movement in the 3-day approvals:
10/5-7 was +7
10/6-8 was +11
10/7-9 was +11
Then, 10/10 approval was not as good. Can't say exactly, but some combination of a really good day falling off, with a less good day on. But 10/11 was worse, and 10/12 was very bad:
10/7 rolls off and 10/10 rolls on, -2 (+11 to +9)
10/8 rolls off and 10/11 rolls on, -3 (+9 to +6)
10/9 rolls off and 10/12 rolls on, -4 (+6 to +2)
10/10 rolls off and 10/13 rolls on, -1 (+2 to +1)
But approval then starts going back up once the very bad days roll off:
10/11 rolls off and 10/14 on, no change (+1 stays +1)
10/12 rolls off, 10/15 rolls on, +3 (+1 to +4)
today,
10/13 rolls off, 10/16 rolls on, stays +4.
There's certainly been a sustained dip in the 3-day approval numbers since the first debate, from +11 to +4, but remember, the worst two days that caused the massive approval hit (10/11 and 10/12) are STILL ON THE 7-day tracker, which is 10/10-10/16.
I suspect the numbers will stay the same tomorrow, and start trending up. Other daily trackers, not on a 7-day window, are already showing this. Gallup is a lagger due to the really long window. But it's going to come up. Slowly if there's no second-debate bounce (horrible numbers off, mediocre numbers on), or faster if there is a second-debate bounce (horrible numbers off, good numbers on).