CCR

[donovan]

.....I am more concerned about people that are turned on by any computer poll.....it takes so little..

[CFP Admin]

Some people would be turned off by a computer poll that picks North Dakota State to lose by 20 when they win by 30. That was about a 50 point swing right there.

First of all, Minnesota was picked by 22 and lost by 6. That's a 28-point swing. Iowa State was favored by 10 over Northern Iowa and lost by 11. You can see that kind of "upset" even among FBS schools every year. But the low percentage of their occurence make these anomalies - and you don't adjust a formula for anomalies.

Second, you seem to be ignoring the point about how small the sampling is of FCS schools vs. FBS schools. My opinion is that you can only pretend to measure a school that plays 2 FBS teams against those that play 10-12 FBS teams.

It is obvious that, if you measured all of the FCS schools against their inter-divisional play, you would probably improve the spread prediction. But you are rarely going to see a match-up where it makes any major difference. The weakness of the other FCS schools would keep the numbers down.

[Spence]

I agree. You can try to factor in everything, but even in D1 you can never project a Stanford over USC result and they play like competition. If I were setting up a rankings system I would never put in the time to rank the 1AA teams unless I were ranking their league as a whole. Putting them in for a few games in 1-A wouldn't be worth the work IMO. Of course I am on the record as believing that the two conferences shouldn't play at all. Even with an upset like AP State over Michigan, these games are a waste of time and a real waste of money for fans who want to see D1A match ups and games that actually tell you something about the teams.

Ap State comes in a beats Michigan then leaves back to their conference. What does that tell us about how Michigan matches up with anyone in D1A? Nothing. Maybe AP State could beat 60% of D1A teams, maybe they could only beat Michigan. They provide no direct matchup comparison. making the game useless IMO.

[donovan]

There is the priniciple that one time does not predict future events. Most of us learned this very early. You walked down the road and picked up a rock and threw it a the telephone pole. You hit it on the first throw. The smartest guy never threw another rock. The rest of us tried to prove something. Same thing.

[Eric]

Actually, I was talking about the ND State/CMU game and I was being generous there. I rounded down, it was actually a 55.something swing.

I know it would take a ton of work to do all of the FCS teams, but couldn't you get the CCR to take into account some of the returning players along with the stats and the current ranking in some of the FCS polls and roll it all up into a weighted average? The least it could do is make the spread not sound so ridiculous.

[CFP Admin]

I play with different ideas all the time. In the offseason, I have replayed entire seasons with alterations to the formula. I'm always curious to see how the rankings could improve from 75% to 80% accuracy on game picks. Or how teams would be ranked if "A" meant more and "D" meant less. Oddly, every time I think I've found an answer, the result either goes in reverse or simply trades wins-and-losses. I don't think that means I've found the perfect formula. It may suggest that I'm too dumb to get to the next level. I never claimed to be the world's greatest mathematician anyway, and there's no one I know that would call me a "geek". I'm just a college football fan who is having fun and sharing information.

[Spence]

There is the priniciple that one time does not predict future events. Most of us learned this very early. You walked down the road and picked up a rock and threw it a the telephone pole. You hit it on the first throw. The smartest guy never threw another rock. The rest of us tried to prove something. Same thing.

I agree that what happens in one game doesn't necessarily tell you anything. If you watch the individual match ups, though, and know the strengths and wweaknesses of a team, you can get a pretty good idea how teams compare. That is the closest you can come to predicting what will happen if two teams play. You can't figure in uncharacteristic turnovers or other chance things that can happen, but over time you can get a pretty good idea how good a team can be. Nothing will tell you exactly what will happen.

[CFP Admin]

Actually, I was talking about the ND State/CMU game and I was being generous there. I rounded down, it was actually a 55.something swing.

You're right. Central Michigan was favored by 25 over NDSU and lost by 30 (44-14).
Central Michigan was also an 8-point Vegas underdog to Kansas in the season-opener. Kansas won by 45.

How do these results prove anything other than CMU was over-rated?? Hmmm, CMU was a 4-point dog at Ball State but won by 20. Was CMU under-rated??

Maybe they aren't under-rated or over-rated. Maybe they are simply inconsistent this season and there's no way you foretell their results with accuracy.

[Eric]

They've sucked in OOC play but not in MAC play. Sounds like that's a trend to me.

[CFP Admin]

I respect you and your opinions. But I was asked to clarify a particular matter related to how the CCR determines spreads of FBS teams vs. FCS teams. I did that. I know from experience that this, and every ratings subject, can be nit-picked to death and come back full-circle to start again, living on in perpetuity. You can all debate it without me. I'm not suggesting there isn't a way to do it. As I've said, I try constantly to find improvements. I may someday even add FCS teams into the equation. Right now, Sagarin and some others do this. I don't. I focus solely on the FBS.

[Eric]

Well, I know that computers aren't allowed to realize that.

I'm not trying to attack the poll, I was just asking about the subject and wondering how it could possibly be improved. I have no clue how the computer knows how to pick games and I'm sure I couldn't do a better job

[CFP Admin]

Well, I know that computers aren't allowed to realize that.

I'm not trying to attack the poll, I was just asking about the subject and wondering how it could possibly be improved. I have no clue how the computer knows how to pick games and I'm sure I couldn't do a better job

I know you're not attacking anything. I'm not offended, even if you do. I question it constantly because I want it to be as accurate as possible.

And don't be so sure you couldn't do a better job.

These things are certainly all open to debate. That's partly, at least, why we have a message board. All I'm saying is I can't dedicate the time to be here and answer every question, or debate every issue. The board is for the membership to discuss matters freely, with as little intervention or involvement from us as possible.

To be more candid with your original question, "Has anybody else noticed that the CCR struggles picking the FCS upset?" It doesn't struggle so much as it simply isn't set-up to really care too much. The computer used to give all FCS teams a rating of "0". With more games being played, I revised the formula to average the results of all FBS vs. FCS games. The purpose of the change was to lessen the severity of the impact on the FBS teams' ratings. Whether or not I go further and actually rate FCS teams, is something I will have to give more consideration. What I always have to be mindful of is that I don't reduce the accuracies by trying to improve the inaccuracies.