Dullard's College Football Ratings - Week 11

[dullard]

Sorry for not replying to this yesterday. I got busy.

also... curious... how is Bowling Green all the way down to 43?

My program is based almost entirely on strength of schedule and how well you do compared to those opponents (margin of victory). My program doesn?t reward teams that play cupcake schedules unless you win big. For example if an average team plays a top rated team, they can soar in the rankings by beating them ? even if it is just a narrow margin of victory. But if a team plays a bottom rated team and only win by a couple of points that can very likely hurt your ratings. Basically if you play cupcakes and barely win ? you aren?t very good yourself. Bowling Green strength of schedule has the #113 out of 117 teams. Meaning they have almost the easiest schedule of any team. Bowling green did not blow most of these teams away. They only beat the team in last place (Eastern Michigan) by 13 ? while almost every other team beat Eastern Michigan by 20 ? 30 points. Performance like that hurts.

Edit: I've also been a bit surprised at how USC is ranked in the BCS. The BCS takes the records of the team(s) you've lost to and calculates them in. USC lost to a team that is
5-5 and they are still ranked in the top 5 in the BCS as well as every other poll.

Your statement is mostly incorrect. Lets look at the BCS components:
1) Quality wins ? a loss to a 5/5 team doesn?t affect this.
2) Number of losses ? a loss to a 5/5 team doesn?t affect this any differently that a loss to a 9/1 team.
3) Strength of schedule ? this doesn?t care which teams you won to or lost to. In fact if you do the math, losing games helps your SOS since then your opponents have more wins and fewer losses!
4) Computer average ? this part depends on which computer you are using, some programs have a 5/5 loss hurt you and others don?t.
5) Poll average ? this 5/5 loss may or may not affect the way voters vote, sometimes they consider it and other times they don?t.
So basically your statement is only partially true in two of the five BCS components (sad isn't it)!

 

[dullard]

USC, FSU, LSU....doesn't matter. The Sooners own them all.

All the voters and computer programs (including mine) said that about Miami last year ? and they lost their only game in overtime at the national championship game.

Do you keep track of your program's record for the past 3 weeks (wins vs. losses, regardless of the error in points). Can you tell that the more games that are played, the better your program gets?

Generally about 82% of the teams are ranked ahead of teams they beat. Generally the average error from subtracting the scores is right around 10 points. The performance of the program is a bit better towards the end of the season (since it has more data), but on bowl games its predictions are horrible. This is since teams often have 6 full weeks to prepare for one game and anything can happen.

 

[dullard]

Eh, I don't really think so. I think OU's schedule has not been as tough as some of the other teams in the top 5. Thus, when they finally play a team that is ranked better than UT or OSU, they might be in for a bit of a surprise.

My program lists OU?s schedule at #18. I get this by averaging the ratings of the teams OU has already played. So yes many teams have had a tougher schedule. But OU?s schedule is harder than ANY of the 1 loss teams ? and tougher than most of the 2 loss teams.

Do you have an easy way to enter in the results from a week's games, or do you manually type everything in?

I go to ESPN.com and copy and paste the scores right in. It does the rest. In the past I?ve also copied and pasted from other sources, but I?ve been too lazy this year. So if ESPN goofed big, I guess I have errors. By the end of the regular season, hopefully I?ll have time to double check every score of every game.

 

[austin316]

out of curiosity, can you list the teams strength of schedules also? or at least michigans and Ohio States?

 

[dullard]

Originally posted by: austin316
out of curiosity, can you list the teams strength of schedules also? or at least michigans and Ohio States?

Do you want them ranked (#1, #2, #3) or do you want the average strength so you can compare the numbers directly? I'll edit my original post when you reply.

 

[sciencewhiz]

How do you rate games against non division 1 teams? Bowling green narrowly beat #20 Purdue, and lost by only a touchdown to #16 Ohio State, so I would have though they would have been a little higher.

I was curious about how you input the scores, not because of the possibility of errors, but because of the time involved. I seem to remeber one of the BCS computers saying that they used data from Peter Wolfe, and so was curious if you had some automated process.

 

[dullard]

Originally posted by: sciencewhiz
How do you rate games against non division 1 teams? Bowling green narrowly beat #20 Purdue, and lost by only a touchdown to #16 Ohio State, so I would have though they would have been a little higher.

I was curious about how you input the scores, not because of the possibility of errors, but because of the time involved. I seem to remeber one of the BCS computers saying that they used data from Peter Wolfe, and so was curious if you had some automated process.

There are companies that sell the data. I don't know if Peter Wolfe gives it out for free or if he charges. I only include teams in Division II that play multiple Division I games. In those cases I hand type what I find on their website if they have a website that is up to date. If not, I don't include them. I've found that the Division II results have absolutely no effect on the top 25 1A rankings - so dropping them isn't very important (since I only really care about those teams).

As for Bowling Green, the ratings are a season average. You cannot take just one or two games, ingore the rest, and use that to rate a team. Sadly all humans operate that way - ingoring most of the season and only focussing on one or two games per team. But lets look at the data you provided. Ohio State is rated 127.2, Bowling Green is 118.7. Subtract the two scores and it says Ohio State should win by 8.5. Ohio State won by 7. I say that is quite a close match, being off by just 1.5 points. Considering that one play can swing the result by 14 points (imagine a touchdown pass that should have been caught, but was bobbled, intercepted and ran to the other endzone for a touchdown), an error of 1.5 is quite good. The Purdue game I think was an upset. Any rating system will not be able to predict all upsets. Purdue should have won by 7.6 according to my program, but Bowling Green won by 1. That is an error of 8.6. Basically it is off by almost one play - a touchdown. However the Purdue game is an example of where Bowling Green did better than expected - there are equal games where Bowling Green did worse than expected. Thus on average Bowling Green gets its so-so ranking. Miami, OH is next for Bowling Green. If Bowling Green wins, they will move up. Lets come back tomorrow and see what happens. The computer predicts Miami to win by 10. So even a loss by a couple will help Bowling Green. But a blowout will hurt Bowling Green.

 

[sciencewhiz]

Yes, the human side of me says that Bowling Green is underanked, whereas Miami-Ohio is overranked. Only big team Miami-Ohio has played is Iowa, and they were blown out. But, as you said, we will see.

That's why I love college football 🙂

 

[dullard]

Originally posted by: sciencewhiz
Yes, the human side of me says that Bowling Green is underanked, whereas Miami-Ohio is overranked. Only big team Miami-Ohio has played is Iowa, and they were blown out. But, as you said, we will see.

That's why I love college football 🙂

The human side is often the correct side. Take last weekend. I debated with everyone in my office - saying that Nebraska has no chance against Texas (I'm a student at Nebraska). My computer said Nebraska should win, but I knew better. Sometimes data the human's know isn't in the computer (ie Nebraska's inability to stop Texas's style of offense or the fact that Nebraska is a one man team and any defense that puts an extra man on Nebraska's quarterback will win).

 

[sciencewhiz]

I've learned not to root against my college (undefeated since 1978).

You said in an ealier thread that you cap MOV. How does that affect USC this week, who should beat Arizona by 40+ points by your rankings. Does that mean that USC is penalized if they beat them by 40, becasue of a MOV cap, or does that mean that it doesn't matter if they beat them by 40 or 70, but would be penalized sligtly if they beat them by 30? Or does it mean that as long as the beat them by the MOV cap (even if it is less then the predicted differential) it doesn't matter.

 

[dullard]

Originally posted by: sciencewhiz
it doesn't matter if they beat them by 40 or 70, but would be penalized sligtly if they beat them by 30?

Basically USC has everything to lose and very little to gain by playing such a cupcake. Lets say the cap ends up at 35 points. If USC wins by 35 or more, it would only get credit for 35 points. Add 35 to Arizona's and you are just about at USC's rating. That means USC cannot move up even with a big win. What would happen instead is that Arizona might move up so that the 35 points is a closer rating, USC cannot move up though. Now if USC wins by 30, USC will get a slight drop, if they win by 25, USC will get a bigger drop. Etc. If USC wins by 1 they will drop a lot. If Arizona wins by 1, USC will plummet. If Arizona wins by 35, USC will fall off a cliff and never survive. Ok that is a bit of an exaggeration. But hopefully that answers your question.

Remember though this is just one game in an average. So if this one game says USC should drop 10 rating points and USC plays 10 games, on average USC will just drop by 1 rating point.

Lets think of the worst case senerio: Arizona wins by 35. That means that USC score was about 76 points too high. Of that, USC would drop 38 and Arizona would rise 38. However with the average, this is just one game of ten. So USC max would drop by 3.8 in the ratings from that game. Now you have to also consider how all the teams that USC played in the past did during the week as well... I could go on and on.

 

[MikeMike]

ive got a question, how do teams who beat someone, miami, and ohio state for instance, stay ranked lower than them? iowa, and FSU are who they beat.

MIKE

 

[dullard]

Originally posted by: nourdmrolNMT1
ive got a question, how do teams who beat someone, miami, and ohio state for instance, stay ranked lower than them? iowa, and FSU are who they beat.

MIKE

I'll answer that with another question. Can you come up with any method to rank all teams higher than ALL the teams they beat? Look at this weeks top BCS for example:
*Cal beat USC, but USC is #2 and Cal is nowhere in the BCS.
*Miami beat Florida State, but Miami is below Florida State.
*VT beat Miami, but VT is below Miami.
*Wisconsin beat Ohio State, but Wisconsin is nowhere in the BCS.
*West Virginia beat VT, but WV is nowhere in the BCS.
*Florida beat LSU, but Florida is below LSU.
*Both of the teams that beat Michigan are below Michigan...

Need I go on? All team rankings will have the situation where you are ranked below teams you beat (except for the #1 undefeated team). There is no possible way of avoiding that problem. Frequently you get situations like this: In 2001 Texas blew Colorado away, then in the Big 12 championship Colorado beat Texas. Now who gets ranked higher? No matter what, someone will be ranked higher than a team that beat them. Or for a more common example what do you do when X beats Y, Y beats Z, and Z beats X? That 3 team triangle happens every weekend. So who do you rank ahead of the others? No matter what you will have a team ranked above the team that beat them!

You need to realize that you cannot avoid the problem you posed. One game doesn't define a team. If a team loses all of its games and then beats Oklahoma, should that team with a 1 and 11 record be #1 and go to the national championships? NO! You look at all games equally and average them. Teams suck some weekends and play flawlessly on other weekends. The average gives you where they deserve to be ranked.

 

[sciencewhiz]

Found Wolfe's data: http://www.bol.ucla.edu/~prwolfe/cfootball/scores.htm He includes Divisions I, II, and III and NAIA, so it's probably more trouble then its worth.

2001 is the year that Colorado beat Nebraska in the last game of the season (BIG), then Colorado narrowly beat Texas in the championship game. But, Nebraska got to play for the national championship, without even winning their conference. To me, that is the most glaring example of where a rating system can't always put the best team in.

 

[dullard]

Originally posted by: sciencewhiz
Found Wolfe's data: http://www.bol.ucla.edu/~prwolfe/cfootball/scores.htm He includes Divisions I, II, and III and NAIA, so it's probably more trouble then its worth.

2001 is the year that Colorado beat Nebraska in the last game of the season (BIG), then Colorado narrowly beat Texas in the championship game. But, Nebraska got to play for the national championship, without even winning their conference. To me, that is the most glaring example of where a rating system can't always put the best team in.

No the BCS isn't perfect. I've explained my preferences in the past and don't want to go into detail again. Nebraska should never have gone with the season they had (didn't Troy State nearly win that year?) even though Nebraska only had one loss.

I'll take a look at that Wolfe data. The key issue is that I have a list that converts all the team names to a number. With each data source I have to manually type each team into another list. But luckilly that is a one time task (with minor additions as teams change each year). For example one website might have it listed as Miami Florida while another uses Miami (FL). So the computer has to be able to recognize all those variations. Maybe by the end of the year I'll have the time to make this list.

 

[sciencewhiz]

I kind of like this explanation from SI.com:

I don't care about "strength of schedule", and I don't care about "quality wins". How can ANYONE justify ranking Miami ahead of Virginia Tech in the upcoming BSC Standings after that major butt kicking?
-- Chris Cornelison, Rancho Cucamonga, Calif.

Chris, I'm no math expert, so this is just a guess, but it might be because the BCS has yet to correct the kink in its standings that, for some unexplainable reason, takes into account a team's entire season rather than just one game. Because of that, when the computers added in Miami's previous wins over Florida State and Florida, it may have had the unintended consequence of keeping the 'Canes above Virginia Tech, even though the Hokies did have big earlier wins against Syracuse and Connecticut.