it's been too long since I've taken statistics, but I take your point. that said, college football seasons are inherently tiny samples, so hard proof is going to be a tough standard. But I think the numbers as they are are pretty damning.
The Vols were 2-3 this time last year too with the exception that UGA and Bama are their next two games this year.
Your sample size argument is valid, and A-Smith would be the guy to talk to about actual analysis, but just saying that one number is larger (or smaller) than another is not proof that they are actually significantly different. When your standard deviations overlap, you can roughly conclude that the two numbers are not statistically different. That's because the real value lay somewhere +/- that average. When the standard deviations overlap, the real value of the first measurement may be exactly the real value of the second measurement. When the deviations don't overlap, there is significance, because if the real value of either measure was at either extreme, there'd still be a difference between the two numbers. As it stands, these deviations overlap, so you cannot say that the real value of either measurement is significantly different from its counterpart. Based just on these numbers, there has been no improvement, on average.
It did, The Coach BJ defenders have just reached climate change denier status with me and I'm hyper sensitive to giving them even a tenth of a yard improvement as fuel for their idiocy.
My only problem with the fl0at-style critique is that by the same token, you can't make an argument that the defense has regressed, which I think it clearly has. With sample sizes this small, you're always going to be short of total proof, but I think the numbers can lean you pretty hard in one direction, even if they can't prove anything. And overall, they're leaning pretty hard in the "meh, nothing much different" direction.
While it won't help with sample size, breaking it up by halves may tell a more complete and palatable story.