(12-15-2018 09:19 PM)bullet Wrote:  

(12-15-2018 07:10 PM)Hokie Mark Wrote:  It's incredible to me how many people seem to think that a committee of fallible human beings can more accurately pick the 8 best teams* than a bunch of conference champion winners... the classic "eye test is better than results on the field" argument.

Look, if you're worried about having some fluky upset team take one of the 8 spots, you can greatly mitigate that by simply requiring all conferences to scrap divisions and match the two highest-ranked teams in the conference CCGs. Winner gets an auto-bid; loser (now with another loss) goes into the pot for possible at-large bid.

* keep in mind the committee's selections are based on their opinion and stats from a 12-game season, in which most of the teams under consideration did NOT play each other and may have very few shared opponents even.

Which is another problem with computers. 12 games is simply not a big enough sample to be valid no matter how many different systems you include. Basic statistics tells you that. 6 pieces of garbage added together still equals a pile of garbage.

That's putting it WAY too strongly. Yes, computers aren't perfect, otherwise all computers would have the exact same results, and they don't. But on the other hand, they aren't random garbage either. E.g., looking at the Massey Composite, we see that the top 5 teams are ... Alabama, Clemson, Georgia, Notre Dame, Ohio State. The bottom 5 are .... New Mexico State, Central Michigan, UConn, Rice, and UTEP.

We can quibble about the order, or whether either list leaves a team or two out, but there's nobody who can doubt with a straight face that the first 5 are obviously among the very best teams and the last 5 obviously among the very worst teams. So there is enough data for the computers to create those very reasonable lists.