BCS Computers Revisited

Last month we discussed here the problems that were created when margin of victory was taken out of the BCS computer formula. We looked at Jeff Sagarin's rankings that are used as part of the BCS formula. Sagarin's rankings are nice to examine because he calculates both his original overall rating and also what he calls ELO-CHESS, which he describes as the "politically correct" rating that is used by the BCS. Scoring margin is not included in the ELO-CHESS rating, only winning and losing matters.

In early October we noted some major differences between Sagarin's ratings and the ELO-CHESS ratings that are used by the BCS. At that time the Northern Iowa Panthers were ranked 56th in his "regular" ratings but were #6 when scoring margin was removed. Obviously it was still very early in the year and we figured things would even themselves out as more games were played and as UNI played more opponents from their FCS Subdivision (Northern Iowa is ranked #1 in the Division 1 FCS Subdivision - formerly Division 1-AA.) Well as far as things working themselves out....they haven't. In the current ratings found on Jeff Sagarin's website, through games of November 10th, Northern Iowa is ranked 35th. Seems a bit high for a FCS Subdivision team, but not completely unbelievable. UNI is 10-0, ranked #1 in their subdivision, defeated Iowa State on the road earlier in the year by 11, and would give many FBC teams a very tough game. But once again when scoring margin is subtracted from the formula, things start getting ridiculous. In the ELO-CHESS rating that is used by the BCS formula, Northern Iowa is ranked at #12. That is to say that as far as the BCS is concerned, UNI is the 12th best team in the entire nation. To put that into more perspective, at #12 UNI is ranked higher than USC (#15), Boston College (#14), and Clemson (#13). Now if someone on the street came up to you and was arguing that Northern Iowa was a better football team than Southern Cal you would laugh and immediately disregard anything they further said. Well, in this case that person is the computer polls in the current BCS system and should be considered useless.

Now this is not Jeff Sagarin's fault or his computer's fault. Sagarin is just doing what he was told to do by taking margin of victory out of his formula. We've argued before that scoring margin needs to be included in the BCS computers and we will continue to do so. For example, shouldn't Texas be rewarded more for beating Iowa State 56-3 compared to UNI's 24-13 victory over the Cyclones? Obviously. Would Texas beat UNI in a game played on a neutral field? Obviously. (Texas is #16 in Sagarin's ratings, UNI is #35. Texas is #25 in ELO-CHESS, UNI is #12.) Now I know that this is a very small scale example, and these computer polls take every game played in the entire system into account. But the point remains that if any part, albeit a small part, of the current BCS system thinks that UNI is the 12th best team in all of college football, then that system is flawed. Step one to fix the BCS, please allow these computers to do what they were originally designed to do and put some sort of measure for scoring margin back into their formulas. It really is common sense.

Enable Comment Auto-Refresher

RawbeezeitzMajor Leaguer

748 days ago

Score 3

You make a great point, but I still don't think scoring margin should be taken into consideration. Take, for example, Boston College's 42-35 loss to Maryland. That score suggests a close game, when in fact Maryland dominated, and BC had to score 2 late touchdowns to make it that close. Or what about teams that leave in their first teams for too long and score points when already up by 49 in the 4th? Does that mean they should be deemed as better than teams that finish off games by running the ball and the clock down?

What could be done is to judge each game a team plays, and give it some sort of score from 0 to 10 (or whatever numbers the BCS people want to use) based on how convincing the win was. It would be a completely subjective thing, but if you had 5 or 6 different people who rated every single game, then averaged it out, subjectivity could be reduced.

Of course, a playoff system would eliminate the significance of all this BCS nonsense.

Permalink | Reply

JamesmoutonWaterboy

748 days ago

Score 2

I agree with most of your arguments. A solution to a team leaving in 1st teamers to "run up the score" would be to put a cap on the scoring margin element. That is to say any win of 21 or more is considered equal. So if a team wins by 21 or a team "runs up the score" to win by 50 it doesn't matter. Now would you have teams up by 14 in the 4th quarter trying to score late to get to that cap? Maybe, but no systems is perfect, as we obviously know! I still think that by taking out margin of victory we have handicapped these computer polls. So either put it back in in some form or get rid of the computers all together.

Permalink

The PipDiv-I Stud

748 days ago

Score 0

A playoff would still involve some form of rankings and it is hard to think that the computeres would be completely removed from any ranking method. A playoff also creates other issues. It is not a magic bullet and would be just as flawed as the BCS. If Michigan wins the BIG 10 should they as a three loss team have an opportunit to play for the national championship? NO way! Design a playoff that cuts them out this year. Without a pile of money being tossed in the big ten's direction, it won't happen. I don't want a 3 loss team pulling upsets in DEC/Jan and winning the national championship. I like 1AA teams pulling upsets in September ruining another school's national championship hopes. That to me is what makes college football exciting. No playoff system will be flexible enough to cover all the possibilities that arise, there is no "fair" solution.

Permalink

RawbeezeitzMajor Leaguer

748 days ago

Score 0

But if it were a 16 team playoff, or a 12 team playoff, wouldn't a team like Michigan have to pull 4 consecutive upsets, all on the road?

Permalink

The PipDiv-I Stud

748 days ago

Score 0

As I said I'd rather have my upsets in sept/oct/nov than in Dec/Jan. And are you really comfortable with adding 3 extra games to the season? At least in name these are supposed to be "student-athletes". 16 teams is ridiculous and over-kill. 8, I could grudgingly deal with an 8 team playoff. I'd hate it worse than anyone hates the BCS, and I would not waste my time watching college football until the conference championship games.

Permalink

RawbeezeitzMajor Leaguer

748 days ago

Score 0

I'd rather have a champion determined on the field in Dec/Jan as opposed to in a computer in Sep/Oct.

Permalink

Davis21wylieMVP

748 days ago

Score 1

Good article. In pretty much every sport, point/run/goal differential is the single best predictor of future winning percentage, b/c the old adage that "the good teams win the close games" isn't as true as "the good teams win the blowouts". However, if coaches knew that margin was a factor, they would begin to run up the score against weaker opponents to increase their ranking, a practice that would begin to detract from the effectiveness of point differential as a predictive tool. So I think some kind of MOV cap needs to be instituted to remove the incentive for running up scores. Perhaps when the margin is >28, we cap it at 28, or maybe we cap it when a team's win probability exceeds a certain %. But I agree that MOV needs to be included to get the best indicator of team strength; we just have to find a way to prevent coaches from abusing a MOV-based system.

Permalink | Reply

RawbeezeitzMajor Leaguer

748 days ago

Score 0

Also, with margin of victory being a factor, schools would schedule opponents they knew they could blow out with ease, instead of other good schools that would provide closer competition. And I think it's impossible to accurately give a numerical value to different kinds of wins. What's the difference between beating some I-AA team by 40 points as opposed to a very good I-A team by 4 points? Which is more impressive, and by how much?

Permalink | Reply

The PipDiv-I Stud

748 days ago

Score 0

Nice aritcle, I like that he posts both rankings. I started looking at the computer rankings individually for the first time this year and I'll need to pay more attention to them in the future. There is a MOV issue, but I have no idea how to fix it without hurting the strength of schedule weighting, which to me matters more.

Permalink | Reply

False ProphetAll-Star

748 days ago

Score 0

Elo chess doesn't work that well in games measured by scoring. It's designed for Chess, where the problems we see with it being used in CFB don't happen.

I don't follow his CFB rankings that much, but his NFL rankings with the same formula are usually pretty good (the combo rankings).

And Pip, if Michigan wins the Big 10, they deserve to make the playoffs. Just like the Seahawks might win their division with a 8-8 or 7-9 record. If your conference allows you to get a birth into the tournament, and you win your conference, you deserve to get in.

Permalink | Reply

The PipDiv-I Stud

748 days ago

Score 1

Yeah, I have an issue with a 3-loss team winning the national championship. It should be about who had the best season, not who "survived their cofnerence" then got on a hot streak. The first time that a 3-loss Notre Dame team makles the playoff people will have BCS level meltdowns. So the Michigan (hypotheital winner of Big-10) makes it to the fianl of the playoff, and plays Oregon who beat an undefeated Kansas (after they beat missouri and Oklahoma) in the Semi-final: in this scenario Michigan is a worthy title contender over kansas because they won two or three games in december and despite losing to 2 unranked opponents, and kansas loses 1 game, which happened to be their third top 5 opponent in just over a month. Kansas (assuming they prove themselves) v Oregon would make a great title game where Oregon v Michigan to me would lack drama and feel artificial. I couldn't watch it without thinking "appalachian st" the entire game. The current system is not perfect, but a playoff makes the regular season irrelevant.

Permalink

CLEBloggerJV Squad

748 days ago

Score 1

I don't believe it'd make the regular season irrelevant at all. People still watch the NFL regular season

Permalink

RawbeezeitzMajor Leaguer

748 days ago

Score 0

Having a playoff system that's based mostly on conference championships would keep the regular season relevant. In fact, it would become relevant to the half of D-IA that starts out with no chance at winning the championship: the non-BCS conferences. If a playoff system is devised that allows only conference champions, and perhaps one or two at large schools, then the conference schedule remains as exciting and important as it is now.

Permalink

False ProphetAll-Star

748 days ago

Score 0

I think that Rawbeezitz is right. I think that the rule needs to be designed so Notre Dame plays a factor moreso, probably if they have 2 or less wins they qualify. Not that they'll need it any time soon :)

Permalink

The PipDiv-I Stud

746 days ago

Score 0

Right the conference schedule is all that would matter. That Oregon-Michigan game would never have been scheduled, what;s the point, the win benefits neither team. But if Michigan beats OSU, Oregon can rightfully claim to have beaten the big-10 champ. Right now that matters and amy be the difference between them and the big-12 champ playing for the title.

Permalink