Over the weekend, my Michigan Wolverines lost to the hated Ohio State Buckeyes. If any pleasure can be taken from the loss, it’s the knowledge that it was a close game and likely winnable, and that OSU won’t get any real benefit from it anyway since they’re disqualified from the Big Ten Championship game and any bowl games. Another benefit is that it gives me an excuse to pile on the BCS and discuss why, even if they move to a playoff system, college football will always have an issue with choosing who the best teams are.
Before Saturday, Ohio State was ranked fourth in the AP top 25 poll; they cannot be ranked in the BCS or coaches’ poll due to their sanctions. It’s possible (although who’s to say) that they would be ranked higher if they weren’t under that cloud; OSU is undefeated, a feat shared only by Notre Dame. On the other side, Michigan was 20th in the AP and 19th in the BCS. The BCS rating came from their standings of 20th in the Harris poll, 20th in USA Today, and 19th in the computer rankings. So, going into the game, you would expect Ohio State to win. Not only were they probably the better team, but the game was in Columbus and Michigan had their starting quarterback playing running back and their starting running back sitting with a broken leg. The line seemed appropriate, making Ohio State about a 4 point favorite.
The game itself seemed to follow expectations. Both teams scored on explosive plays, but otherwise the two defenses were pretty stout. Michigan took a one point lead into halftime, but gave up two field goals in a defensive stalemate of a second half to lose by five. In short, it was pretty much exactly what you might expect.
This is where the issue with the BCS comes into play: what should happen if everything goes exactly as you would predict? I think most people would say you should keep the status quo. If you were Brady Hoke, or Urban Meyer, would you be particularly disappointed or elated with the outcome? As a fan, do you think your team is any better or worse than you did before the game? It’s hard for me to see why you would. Had Michigan won, or had they lost spectacularly, you could make an argument that someone was improperly rated. One team might be better or worse than we thought, or even then they might be who we thought they were and something flukey happened during the game. But none of that happened; the favorite won at home by pretty much what we expected them to win by.
What happened in the polls? In the AP, Ohio State stayed at 4th and picked up two points. Michigan, however, fell to 21st and lost about 60 points. The same thing happened to Michigan in the Harris and USA Today polls; they fell to 23rd and 24th. However, Michigan kept their spot at 19 in the BCS. Why? Because they actually moved up in the computer rankings. And this should make something that we all already knew pretty obvious: people rankings are about losses with very little context while the computer polls take context into account. Voters saw that Michigan lost, and even though it was a decent and expected loss they knocked Michigan down the list. The computers looked at Michigan’s performance and who they played and decided that they had done pretty well for themselves. Coupled with some nearby teams doing less well (Florida State, Texas, UCLA, for example), Michigan managed to improve their relative position.
This is, in short, the problem with the BCS: people are in charge of it. We know that coaches don’t watch every football game, and that sometimes they don’t even vote in the coaches’ poll. The people that vote in their place don’t watch every game. Even if they did, they would have incentives to vote certain ways instead of objectively. If I were Brady Hoke, I would be voting up every team I played against to improve my apparent strength of schedule. The Harris Poll doesn’t use anyone currently associated with a team (as far as I know), but that certainly doesn’t mean that it eliminates any bias. And I still doubt that they watch all the games. Even if they would if they could, no one has time to; the polls are updated the next day! If you took out all the dead time between plays, I’m not sure you’d have time to watch the twenty-plus games you’d need to in order to see all the teams that might get votes.
Computers, on the other hand, “see” every game. The information may be decontextualized in terms of when a catch was made or if an interception was thrown on a late-game Hail Mary, but all the information is there (except for one of the most reliable pieces, margin of victory, which the BCS has required to be eliminated from computer polls). It can process all that information quickly, since it’s a computer, so it isn’t a problem to generate ratings for the next day.
Perhaps more importantly, computers know how to weight the information in a way to better rank teams. In the NFL, for example, people seem to think that having a strong run game is important. However, the last few years have shown us that good teams can have apparently weak run games as long as they pass well and run in good situations (when they only need a couple yards, or when runs are unexpected). It seems reasonable to expect that people would overrate teams with good run games, all else being equal, whereas the computers would not. Additionally, there are six computer ratings used to generate one BCS ‘vote’, so even if you think one has poor weightings presumably the others work out to help cover it up. Biases in human voters, on the other hand, are potentially less likely to even out just because there are a number of voters.
So, even next year when we have a playoff system, the BCS is still going to have a problem. The problem is that the teams selected to go to the playoff will be chosen at least in part by people. Those people don’t have all the information needed to properly rank teams, and even if they did it’s unlikely that they would use it accurately. They even have explicit incentives to ignore the ratings where they can vote to make their own team look better. Even if we taught them to overcome those biases and to evaluate teams more fully, they wouldn’t have the time to do it with the current media demands. Instead we have a system built almost entirely on wins and losses and a preseason poll.
I don’t know that I really have a solution for the problem. One would obviously be to use only computer ratings, but that seems like a pipe dream. A good number of people, especially those close to sports, are distrustful of computer or math-based ratings. It’s hard to envision human voters being eliminated from the system entirely. A potential solution is to have a larger playoff system, such that enough teams are admitted that those left out have little claim to being the best anyway. However, even then football games are noisy enough that it’s certain that too few games would be played to truly pick the best team that year. So instead, let me just make my own vote: Michigan is the best four-loss team ever. The four teams to beat them have a total of 3 losses, and one of those was to one of the other teams (Nebraska lost to Ohio State). Those four teams are ranked 1, 2, 4, and 14 in the AP poll; they’re 1, 2, 12, and disallowed from the BCS. None of the losses were at home and two were by a touchdown or less. It was a heck of a season, Michigan. I hope you don’t blow your bowl game.