follow me on twitter

Friday, November 5, 2010

College Football 101: why computer rankings aren't really that objective either

Once upon a time, in the college football world, the major conferences got together and decided upon a format that would take the human element out of deciding who would play for the mythical national championship.  They called this format the Bowl Championship Series (BCS), with the idea being that the teams ranked nos. 1 and 2 in the final BCS ratings, would meet in the national championship game.

At the inception of the BCS, the top teams were determined by averaging a series of polls conducted by various media outlets, with the no. 1 and no. 2 rated teams having the highest average of these polls.  Strength of schedule was also factored into these ratings, as were rankings based on computer algorithms, but these components did not nearly carry as much weight as the human media polls.  Overall record played a role in the final BCS standings, as teams who had one or more losses were penalized for each loss on their schedule.

In 2005, the BCS championship would be decided in a game where undefeated USC defeated previously unbeaten Oklahoma for that year's mythical national championship.  There were many people, however, that felt the University of Auburn, who would also finish the season undefeated at 13-0, deserved to play in the national championship.  They argued that Auburn's strength of schedule was tougher than Oklahoma, and thus Auburn would have given USC stiffer test than the 55-19 blowout that the Trojans eventually administered to the Sooners.  They would further argue that the flaw in the BCS system was allowing human media polls to determine who would play for the national title, and the way to remedy this was to find a more objective way to determine who the top two teams in the nation were.

The BCS would respond by eliminating several of the human polls and revamping its ranking process.  In the new BCS rating system, only two human polls(the USA Today Coaches Poll and the Harris Interactive Poll) would be factored into the equation, along with six independently conducted computer-based rankings.  The top teams in the nation would then be determined by adding up each of the polls and then dividing by 100.  The two teams with the highest percentages would then play for the national title.

Fast forward to today, where the recent BCS ratings have TCU, Boise State, and Utah as numbers 3, 4 and 5 in the current rankings, despite playing weaker schedules than many of the teams below them.  Yes these schools are undefeated, and yes all of the schools ranked lower than them have at least one loss.  However, records don't tell the whole story in this case.

Take Boise State for example.  The Broncos are ranked anywhere from no. 4 to no. 13 in the computer polls.  That is quite a discrepancy.  For comparison sake, let's then look at Nebraska and Oklahoma, who at nos 7 and 8 are three and four spots below the Broncos, respectively.  Nebraska is ranked anywhere from 4 to 10 and Oklahoma is ranked anywhere from 4 to 11.  Both examples show a range of rankings but not a near a leap as it is from 4 to 13.  Why is there such a gap in the computer rankings of Boise then? That's easy: each computer ranking code is written by a different individual, who in turn places different weight on factors such as strength of schedule, quality wins, etc.  Wait a second, how is that any more objective than the media polls? There's the rub.  But there's more.

Utah, ranked number 5, does not currently have any wins over teams ranked in the BCS top 25 yet is ranked above several one loss schools that have multiple wins against the BCS top 25. What makes it even worse is that no poll (human or computer) has Utah ranked higher than number 6.  So Utah is rewarded because it has won all of its games playing against easier competition? Sound familiar? It should, because it is the exact scenario the BCS committee was trying to avoid when they created the new system.

So where does this all this fuss about computers leave the current state of the BCS? Imperfect, just like it was before, and, before the dust settles on this season, some team will win the national title while another team cries it was robbed.  One thing is for certain, fans of the current BCS ratings hoping that the computers would shed some objectivity on the situation have to be disappointed.

No comments:

Post a Comment