This is what happened on back when "we" concocted our own computer poll (with ftbobs). It was simply, and goofy, at first, so "we" modified it. Why? To make it come out more like human polls, or what we thought was right.
This is often the curse of computer models.
Mr. PacTen used to have a computer model poll as well, which was as good as anything else, for what it was trying to accomplish.
Pick a philosophy of how teams should be rank-ordered, justify it, figure out how to implement it, and then stick with it. Six different BCS computers had six different methodologies with six different results, largely because the philosophies of the designer were different. And that was fine.
I don't even mind some tweaking...most things need tweaking as you test them for validity and realize a model has deficiencies that aren't hitting what you're aiming for.
But fundamentally changing the goal of the model only undermines its own usefulness and admits that your philosophy must have been crap to start with. Which was the problem with the overall BCS formula. Every time it did something the pollsters didn't like, they changed it so that it would've come out like the pollsters wanted. Why even bother with a BCS formula, then? It would've been more honest to just have polls.
Which is basically what we have now in the playoff committee, whose methodology, best as I can tell, is 1) Whatever we feel like, 2) Whatever makes for the most entertaining tv draw.
Well.....
now we're gonna have 12 playoff teams, which combined with NIL and xfer portal means we just have another NFL. Yay.
