Rage Against the Machines: BCS Ballot Busting, PLUS! Introducing the Computer Wisdom Watchdog
by SMQ on Oct 16, 2006 12:04 PM EDT
in News
Email
PrintThe first BCS standings, as you know, are out:
1. Ohio State (.973)
2. USC (.956)
3. Michigan (.934)
4. Auburn (.748)
5. West Virginia (.745)
6. Florida (.738)
7. Louisville (.718)
8. Notre Dame (.694)
9. Texas (.684)
10. Cal (.668)
Gnash, gnash, wail and tear, but unless you happen to object to the idea of any poll of any variety determining "championship" participants, you don't have much to squawk on about.
You could rage against the machines: the above numbers of course are calculated in large part using the aggregate wisdom of six computer polls, one of which, that of Jeff Sagarin, currently ranks six I-AA schools in its top 60 - all ahead of No. 63 Miami of Florida. Wes Colley's "bias free" method ranks Rutgers seventh and California six spots ahead of Tennessee. Kenneth Massey's input ranks San Jose State at least ten spots higher than Georgia, Virginia Tech, Florida State and Miami of Florida. SMQ would say that result renders the input arbitrary and inconsequential, and therefore not to be accepted to rank the top five, either, but that would make him a Philistine of Science or something. Computers can't be very objective if they produce widely varying results based on possibly meaningless statistical inputs that aren't even allowed to recognize the difference between Tennessee's should-have-been-larger 17-point win over Cal and its one-point win over Air Force.
Penn State ahead of Arkansas? Washington State in the top 25? The Brooklyn Dodgers? Jeff Sagarin's got some 'splaining to do...
But the BCS amalgam is not designed to be the objective end-all of strength ratings; it exists strictly to create a contest its corporate sponsors may dub "The National Championship Game" with the least amount of disagreement and the maximum TV ratings, and by this measure, is the poll en route to invalidity? It is not. As it stands, the Ohio State-Michigan winner will play Southern Cal. If USC does not lose, there will be no objection to this, and BCS backers will cackle with justified glee as the process "works" for the second straight year. Wow! Even if SC loses, the Louisville-West Virginia winner is going to certainly rise high enough with that game and dangerously presumptive wins over Rutgers and quietly surging Pittsburgh to mollify widespread criticism.
Only if there is no undefeated SC or Big East team (that discussion for now ought to include Rutgers, too, which several of the computer polls have in the top ten or twelve, sometimes ahead of U of Hell, and which will rise itself if it beats its more ballyhooed Big East mates), will we have the desired madness, beloved chaos, the Whammy in the BCS' spin on "Press Your Luck," the coveted one-loss toss-up guaranteed to finally spur the hordes into action for the glorious playoff revolution. Or at least a slow morphing of the already expanding big money series into a playoff that accomodates most of its current structure.
That speculation includes Auburn, Florida, Notre Dame and Texas, and if California and Tennessee are going to be part of that mix, it should also include current No. 12 Clemson and, depending on the remainder of the season, Arkansas and Oregon. Any of these teams winning out - especially Cal, Oregon or Notre Dame, which would mean at least one SC loss, though one that would not necessarily drop the Trojans from the second spot, depending on how the Big East shakes out and the ultimate value of SC's "Arkansas Blowout" stock - would mean a random grab for the opponent of the Ohio State-Michigan winner in the mythical championship game. Under that scenario, which includes at least four probably equally deserving one-loss teams, the loser of Ohio State-Michigan would have as good a claim as any of the others, a la Florida in 1996, when its championship came via a rematch with the Florida State team that had beaten it and its eventual Heisman winner to a pulp in the regular season finale. This is unlikely because of the tendency of human polls to overreact to recent losses compared to old ones (Cal is the main beneficiary in that case), and even in 1996 Florida only got its shot because undefeated Arizona State was locked into the Rose Bowl; that the same scenario under the system put in place just two years later would have matched FSU with the Sun Devils, and likely resulted in a clear Florida State win and championship (or otherwise an Arizona State championship), with Florida playing a random game for no higher than number two, is good evidence to the inherently arbitrary nature of all this polling to begin with, and its utter inadequacy for determining a legitimate champion.
2 comments
|
0 recs |
Do you like this story?
Comments
Rage FOR the machines!
There are two interesting situations in the (AP) top ten where the order depends upon how you tally the votes. Southern Cal is third by both Borda (1465) and Bucklin (35 of 65 voters have the Trojans at least as high as third.) West Virginia is just six points behind in points and 33 of 65 voters have the Mountaineers third or better. But if you look ballot by ballot, 33 voters had WVa higher than USC and 32 had the Trojans higher. This could be very interesting if we were talking about #2 vs #3 in the final Harris and/or USA Today polls.
The other interesting situation involves the #7/8/9 teams, Tennessee (1179 points), Auburn (1176) and Florida (1143). Now, Florida beat Tennessee so that doesn't look right. And sure enough, 36 voters have the Gators ranked higher than the Vols, 54 have Auburn higher than Florida, but Tennessee is ranked higher than Auburn 34 to 31. This is a case that Bucklin handles a little better - 33 voters had Auburn but neither of the other two as high as seventh. Majorities had Tennessee and Florida both as high as eighth. If the first tiebreaker for Bucklin is head-to-head, the order becomes [7] Auburn, [8] Florida 36 [9] Tennessee 29, which is consistent with the game results.
As to the computers, the seemingly odd results are just because it is still too early. Through Saturday's games, division 1A was just barely connected by the game vs graph. Only about 25 percent of all possible team-pairs are connected by no worse than an opponents' opponent relationship.
The diameter of the games graph is 5, and average distance between teams is 3. By the end of the year, 50 percent of all team-pairs will be within opponent's opponent of each other, less than 5 percent will be as far as 4 apart, and the average distance between any two teams will be less than 2.5.
Oh, and 'splainin for Jeff
(More Evidence Regarding the Latter: In AP poll #7, eight voters left Arkansas out of their top 25 after a 27-10 domination of then #2 Auburn. In poll #8, all 65 voters had Arkansas no worse than 22nd after a 63-7 over a 1-AA team. Guess some folks needed convincing, and are easily convinced.)
Certain photos copyright © 2012 by Associated Press or Getty Images. Any commercial use or distribution without the express written consent of Associated Press and Getty Images is strictly prohibited.
