Rankings by Top Wins

[gkIrish]

So I was bored and decided to see where teams would fall if you only considered each team's top 3 wins. I only evaluated the 25 teams ranked in the final AP poll plus Texas. My criteria was as follows:

1. Only wins against teams with at least 5 AP votes count.
2. A team's rank = points. For example, Auburn finished 2nd in the ranking so a win over Auburn is worth 2 points. A win over Texas Tech is worth 30 points because they were the 5th team in the "others receiving votes."
3. If you don't have win that qualifies, you get 36 points for each top win missing
4. The lower the score the better. Add the total of your 3 best wins and you get a final score.

The theory behind this is that we are always trying to figure out if a team is overrated or underrated, better than one team or another. But IMO, it always comes down to WHO HAVE YOU BEAT? Notre Dame fares very well:

1 Auburn (30)
2 Florida St. (33)
3 South Carolina (35)
4 Notre Dame (43)
5 Oklahoma (44)
6 Stanford (45)
7 Michigan St. (49)
8 LSU (52)
9 Missouri (59)
10 Arizona St. (63)
11 UCF (64)
12 Alabama (65)
13 Baylor (70)
13 UCLA (70)
13 Texas (70)
16 USC (73)
17 Oklahoma St. (77)
17 Oregon (77)
19 Clemson (79)
19 Pittsburgh (79)
21 Texas A&M (83)
22 Ohio St. (90)
23 Vanderbilt (103)
24 Wisconsin (104)
25 Washington (107)
26 Louisville (108)
26 Duke (108)

 

[irishog77]

Good stuff. Thanks for sharing. Reps.

I wonder if you reversed the point total (ascending instead of descending), how would that effect teams 16-25 (or lower)?

 

[Rhode Irish]

If a team doesn't have enough wins to qualify, they should get 31 points for each win they are short. Wouldn't totally solve the problem but would normalize the data somewhat.

 

[gkIrish]

If a team doesn't have enough wins to qualify, they should get 31 points for each win they are short. Wouldn't totally solve the problem but would normalize the data somewhat.

Good idea. But why 31? Under my criteria, 35 teams have at least 5 votes so shouldn't I give teams falling short 36?

 

[Jebediah Springfield]

Does WHEN you beat team matter?

Like early on in season vs late in season.

Edit: How is Stanford lower ranked than ND when they beat us??

 

[wizards8507]

I think the discrepancy between this methodology and our actual final ranking speaks to our inconsistency, a problem that many of us have observed all season. We can beat good teams but for whatever reason we can't always handle the "sure thing" games (Pitt).

 

[gkIrish]

Does WHEN you beat team matter?

Like early on in season vs late in season.

Well it does matter in the grand scheme of things. But I'm not capable of taking that into account mathematically.

 

[Rhode Irish]

Good idea. But why 31? Under my criteria, 35 teams have at least 5 votes so shouldn't I give teams falling short 36?

Yes my bad. I think your reference to TT in your original post made me assume 30 was the most points associated with any win. But whatever that number is, I'd add one to it. It isn't perfect but would help illustrate the point, I think. (Which, BTW, I think is a very good one. Good job with this.)

 

[gkIrish]

Yes my bad. I think your reference to TT in your original post made me assume 30 was the most points associated with any win. But whatever that number is, I'd add one to it. It isn't perfect but would help illustrate the point, I think.

I made the adjustment. Definitely makes more sense. UCF and Oregon got bumps while USC fell a bit.

 

[Irish#1]

Does WHEN you beat team matter?

Like early on in season vs late in season.

Edit: How is Stanford lower ranked than ND when they beat us??

I think the discrepancy between this methodology and our actual final ranking speaks to our inconsistency, a problem that many of us have observed all season. We can beat good teams but for whatever reason we can't always handle the "sure thing" games (Pitt).

Okay, who's the resident Calculus major on this board? We need some help formulating variables into this equation.

 

[Rhode Irish]

Should send this to all the AP voters. I'd say send it to the coaches, too, but most of them are probably too dumb to get it or too biased to care.

 

[gkIrish]

Should send this to all the AP voters. I'd say send it to the coaches, too, but most of them are probably too dumb to get it or too biased to care.

I have it saved on excel.

I added Texas because that's the only team I could think of outside the top 25 that made sense.

It's amazing that 2 of our 4 losses came to teams ranked 5th and 6th.

 

[RuntheBall]

Interesting. What else could be used, any ideas?

Now how about worst loses? Average the two? And maybe weight for early/ late season, home/away games, and point difference?

 

[woolybug25]

Seems like three games might still skew a good year for many teams. The SEC for instance, usually plays 3 or so really tough games, but play powderpuffs every other week. Seems like this doesn't put anything in for teams that play tough teams week in and week out.

For instance, FSU is greatly helped by winning against Auburn, because they only received one point for it. Meanwhile, their other two best wins equal 32. So if you took Auburn out and replaced it with the average of the other two, they would have a 48. Which isn't nearly as good.

So teams that win their bowl games, get a huge jump in their ranking. But teams like Stanford, who beat 7 ranked teams (went 7 out of 8), don't get the bump because all of the teams they beat ended up in the middle of the pack. If you took out Stanford's lowest score from their "top three" and replaced it with a win against MSU, their number would be 28 (which would be #1 on the board). They are being punished because they didn't win their bowl game, but not getting credit for beating 7 really good teams.

 

[gkIrish]

Seems like three games might still skew a good year for many teams. The SEC for instance, usually plays 3 or so really tough games, but play powderpuffs every other week. Seems like this doesn't put anything in for teams that play tough teams week in and week out.

For instance, FSU is greatly helped by winning against Auburn, because they only received one point for it. Meanwhile, their other two best wins equal 32. So if you took Auburn out and replaced it with the average of the other two, they would have a 48. Which isn't nearly as good.

So teams that win their bowl games, get a huge jump in their ranking. But teams like Stanford, who beat 7 ranked teams (went 7 out of 8), don't get the bump because all of the teams they beat ended up in the middle of the pack. If you took out Stanford's lowest score from their "top three" and replaced it with a win against MSU, their number would be 28 (which would be #1 on the board). They are being punished because they didn't win their bowl game, but not getting credit for beating 7 really good teams.

Perhaps I'll do another evaluation tomorrow for the regular season only. But I do think bowl games matter, especially when you think about where teams should be ranked next season. I thought 2 games wasn't enough because then teams with powderpuff schedules would be rewarded. 4 games was too many because there are very few teams that have 4 wins over top 35 (maybe only Stanford). So yes, Stanford would look better if you used more games but it would be relatively pointless overall.

 

[Irish#1]

Seems like three games might still skew a good year for many teams. The SEC for instance, usually plays 3 or so really tough games, but play powderpuffs every other week. Seems like this doesn't put anything in for teams that play tough teams week in and week out.

For instance, FSU is greatly helped by winning against Auburn, because they only received one point for it. Meanwhile, their other two best wins equal 32. So if you took Auburn out and replaced it with the average of the other two, they would have a 48. Which isn't nearly as good.

So teams that win their bowl games, get a huge jump in their ranking. But teams like Stanford, who beat 7 ranked teams (went 7 out of 8), don't get the bump because all of the teams they beat ended up in the middle of the pack. If you took out Stanford's lowest score from their "top three" and replaced it with a win against MSU, their number would be 28 (which would be #1 on the board). They are being punished because they didn't win their bowl game, but not getting credit for beating 7 really good teams.

Perhaps I'll do another evaluation tomorrow for the regular season only. But I do think bowl games matter, especially when you think about where teams should be ranked next season. I thought 2 games wasn't enough because then teams with powderpuff schedules would be rewarded. 4 games was too many because there are very few teams that have 4 wins over top 35 (maybe only Stanford). So yes, Stanford would look better if you used more games but it would be relatively pointless overall.

I think we still need that math guy. Anyone out there?

 

[ryno 24]

I also think who you lose to is very important as well. College basketball is analyzed very well with KenPom. Efficiency rates are very important. Overall a good analysis though.

 

[RuntheBall]

How about this:

Weighting the 12-ish games teams play in 8% increments.
For wins:
The worst team's ranking (125-ranking) would be multiplied by 8%, then 16% for the second worst team, etc.

For loses the "points" turn negative, and the percentage flips. The worst team played would be ~100% weighting and the best team about 8% weighting.

This would be a way to 1.) use all of the games played 2.) weight games based on opponent strength 3.) Addresses some of the concern's Wooly had.

It would be pretty tim intensive though.

Another weighting could be used to weight for time in the season the game occurred. One way to do this could be a multiplying the first game played by .89 the second my .90 the third by .91 etc. and the final game being multiplied by 1.00

 

[Rhode Irish]

I mean....the beauty of the original thing was that it was simple and helped demonstrate a simple point that is being overlooked. It wasn't a graduate thesis.

 

[woolybug25]

I mean....the beauty of the original thing was that it was simple and helped demonstrate a simple point that is being overlooked. It wasn't a graduate thesis.

I think the point was a simple ranking on "Who Have You Beat". I wasn't trying to complicate it, but that is a trickier question than it seems. Some teams play a few really tough games and nobody outside of those games. While some teams play really good teams, but not elite, almost every week.

Which do you think is more difficult?

A) Beating the #2, #12 and #14 teams in the country. Meanwhile, not playing a single other team ranked better than #46.

B) Beating the #12, #15, #17, #19, #24 and #25th best team. Meanwhile, not playing a single other team ranked worse than #46.

In A, you get 28 points and in B, you get 112 points. Saying both teams are 12-0, the team that only played 3 even remotely difficult games is the best in the country. The team that beat TWICE as many ranked teams and had a much more difficult schedule, isn't even ranked and looks 4x as bad as Team A.

If you had to pick between the two undefeated teams for a slot in the playoffs… I think team B earned the right over team A.

 

[gkIrish]

I think the point was a simple ranking on "Who Have You Beat". I wasn't trying to complicate it, but that is a trickier question than it seems. Some teams play a few really tough games and nobody outside of those games. While some teams play really good teams, but not elite, almost every week.

Which do you think is more difficult?

A) Beating the #2, #12 and #14 teams in the country. Meanwhile, not playing a single other team ranked better than #46.

B) Beating the #12, #15, #17, #19, #24 and #25th best team. Meanwhile, not playing a single other team ranked worse than #46.

In A, you get 28 points and in B, you get 112 points. Saying both teams are 12-0, the team that only played 3 even remotely difficult games is the best in the country. The team that beat TWICE as many ranked teams and had a much more difficult schedule, isn't even ranked and looks 4x as bad as Team A.

If you had to pick between the two undefeated teams for a slot in the playoffs… I think team B earned the right over team A.

I think at some point the quality of the teams is negligible. For example, is there a meaningful difference between the 46th ranked team and the 57th ranked team? The 86th and 101st? I don't know where you draw the line but I just picked the top 35 because those are the teams that enough people thought were good enough to be included in the AP Poll. Your point is well taken though. Stanford probably had the most impressive resume in the country but this is a simplified way of looking at things.

 

[irishog77]

I think at some point the quality of the teams is negligible. For example, is there a meaningful difference between the 46th ranked team and the 57th ranked team? The 86th and 101st? I don't know where you draw the line but I just picked the top 35 because those are the teams that enough people thought were good enough to be included in the AP Poll. Your point is well taken though. Stanford probably had the most impressive resume in the country but this is a simplified way of looking at things.

I think you're right here. But there is a meaningful difference between the 36th ranked team and the 109th ranked team.

That's what baffles about most of the computer rankings. I-AA teams are slotted one spot below the last I-A team. I don't know how many I-A teams there were this year, but if it was 125, then all other teams below I-A are considered 126th. Which we know isn't true. There's certainly even a pecking order in lower divisions.

 

[RuntheBall]

Sorry for over complicating. I agree the metric used was nice clean and simple. I just wanted to see if you could tweak it slightly to get a more in-depth calculation that more accurately reflects SOS and good wins/bad loses, etc.

In the end I just thought multiplication by some weighting factors would be neat. It does complicate it a bit, but using some programming wouldn't be that bad after ranking the teams. In the end it just multiplication, not an SVM classifier based on 3-4 characteristics (thats annoying and super annoying).

Disclaimer: I'm a nerd with numbers and classification/regression analysis. Sorry, in retrospect, I got a bit to excited. I'd still like to run my numbers (even if its not on every team; maybe USC, A&M, ASU,ND and a couple others). Does anyone know a ranking that has all FBS teams ranked 1-125 (I think 125 is the number?). Thanks for any help and apologies for hijacking.

 

[ozzman]

another wrinkle to think about. do you differentiate for home/road?

 

[Rhode Irish]

I think the point was a simple ranking on "Who Have You Beat". I wasn't trying to complicate it, but that is a trickier question than it seems. Some teams play a few really tough games and nobody outside of those games. While some teams play really good teams, but not elite, almost every week.

Which do you think is more difficult?

A) Beating the #2, #12 and #14 teams in the country. Meanwhile, not playing a single other team ranked better than #46.

B) Beating the #12, #15, #17, #19, #24 and #25th best team. Meanwhile, not playing a single other team ranked worse than #46.

In A, you get 28 points and in B, you get 112 points. Saying both teams are 12-0, the team that only played 3 even remotely difficult games is the best in the country. The team that beat TWICE as many ranked teams and had a much more difficult schedule, isn't even ranked and looks 4x as bad as Team A.

If you had to pick between the two undefeated teams for a slot in the playoffs… I think team B earned the right over team A.

I agree that the endurance test of playing 12 real teams deserves respect, I just think what it says about your team is a little different than who the best teams are that you beat. The former says that your team is focused and disciplined and brings it every week. All things that should be considered when thinking about who the best team is. The latter, though, says something about your ceiling.

You could play 12 Purdues or 12 Pitts or 12 Boston Colleges, and if you beat them all that proves something, but it doesn't prove where you stack up against other elite teams.

My honest opinion is that ND's schedule is always one of the toughest, because we test our endurance and our ceiling. Or at least we try (not our fault when Michigan, USC or Texas is down when we schedule them years out). Personally, I value "who you beat" slightly more than "how good was the worst team you played," but I recognize the value of both.

 

[irishog77]

Boise State played the "play 2 tough games all year" crap schedule-- first game of the season and the bowl game. And they obviously did pretty well using that method.

But there's not one legitimate football mind who would say that Boise State would beat Oregon, Oklahoma, or Georgia, week in and week out, over the course of 13 games, 6 or 7 times.

Hell, look at alabama this year-- they scheduled themselves the best possible way prior to the season for their 2 toughest perceived games -- a&m and lsu-- with tune ups or byes.

I think it is tougher to be tested (somewhat) each week than it is to be tested hard a couple times a year...particularly if you can schedule around it accordingly.

 

[Rhode Irish]

You can only "schedule around it" if you are good enough to win the game, though. If you aren't good enough, it doesn't matter who you play the week before. The thing is, being tested every week is hard, but if you don't play and beat any top teams then it is a lot easier to argue that they would have beat the same teams you did than it is to argue that you would have beat them.

Not sure I get the Boise thing....they were 8-5 and didn't play or beat anyone good.

 

[nin05]

I sort of did the same thing throughout the season. I put emphasis on point differential vs. strength of schedule, wins, top 10 wins and top 25 wins. I won't go into the formulas or anything but this is what I ended up with for a top 25. This was done after Army/Navy so there are no bowl games credited to any teams.
1 Auburn Tigers
2 Stanford Cardinal
3 Arizona State Sun Devils
4 Florida State Seminoles
5 Oregon Ducks
6 Oklahoma State Cowboys
7 Missouri Tigers
8 Washington Huskies
9 Alabama Crimson Tide
10 Ohio State Buckeyes
11 UCLA Bruins
12 USC Trojans
13 South Carolina Gamecocks
14 Baylor Bears
15 Michigan State Spartans
16 Georgia Bulldogs
17 LSU Tigers
18 Wisconsin Badgers
19 Clemson Tigers
20 Arizona Wildcats
21 Texas A&M Aggies
22 Ole Miss Rebels
23 BYU Cougars
24 Notre Dame Fighting Irish
25 Georgia Tech Yellow Jackets

 

[woolybug25]

I sort of did the same thing throughout the season. I put emphasis on point differential vs. strength of schedule, wins, top 10 wins and top 25 wins. I won't go into the formulas or anything but this is what I ended up with for a top 25. This was done after Army/Navy so there are no bowl games credited to any teams.
1 Auburn Tigers
2 Stanford Cardinal
3 Arizona State Sun Devils
4 Florida State Seminoles
5 Oregon Ducks
6 Oklahoma State Cowboys
7 Missouri Tigers
8 Washington Huskies
9 Alabama Crimson Tide
10 Ohio State Buckeyes
11 UCLA Bruins
12 USC Trojans
13 South Carolina Gamecocks
14 Baylor Bears
15 Michigan State Spartans
16 Georgia Bulldogs
17 LSU Tigers
18 Wisconsin Badgers
19 Clemson Tigers
20 Arizona Wildcats
21 Texas A&M Aggies
22 Ole Miss Rebels
23 BYU Cougars
24 Notre Dame Fighting Irish
25 Georgia Tech Yellow Jackets

Yikes… bowl season certainly made those rankings look silly.

 

[nin05]

Yikes… bowl season certainly made those rankings look silly.

Sure did. lol Might have to tweak some numbers for next season!