NCAA 1-A Week 12 Beatpaths Top 25

Here’s the Beatpaths Top 25 for just before Week 13 of the NCAA Division 1-A.

Again, grain of salt – for now, I’m using NCAA to try out some new algorithm variants. This one is based on beatpower, but it attempts to drag teams a bit closer to the average strength of the teams in their beatloops. Last week UTEP was unnaturally high – this sort of thing will penalize teams more if they’re in a loop with crappy teams – or reward them if they’re in a loop with teams better than their own rating.

Rank Team Notes BCS BeatPower

1

USC (Beat California) Undefeated means a perfect beatpower (although it’s not the only way to have perfect beatpower).

1

100.0

(64/64 – 0/64)

2

Texas (Beat Kansas) And USC and Texas are the only two teams with perfect beatpower this week.

2

100.0

(70/70 – 0/70)

3

Alabama (Lost to LSU) Seems odd that they’re still ahead of LSU, but LSU does have an extra beatloop team.

8

98.6

(36/37 – 0/37)

4

Oregon (Beat Washington St) Any good arguments for why they shouldn’t be so high?

10

98.0

(50/51 – 1/51)

5

LSU (Beat Alabama)

5

98.4

(61/63 – 0/63)

6

Penn St (Bye)

4

98.4

(59/61 – 0/61)

7

Ohio State (Beat Northwestern)

7

96.6

(56/58 – 2/58)

8

UCLA (Beat Arizona St)

11

98.4

(61/63 – 0/63)

9

Stanford (Beat Oregon St) Stanford gets pulled up due to the high quality of the teams they’ve beatlooped. (The teams they have beatlosses to are already ranked above #9.)

94.8

(45/48 – 2/48)

10

Texas-El Paso (Bye) The other polls have them creeping into the top 25. We had them pretty high last week, but their poor beatloop quality drags them down a bit.

99.1

(54/55 – 0/55)

11

Miami FL (Beat Wake Forest) Not sure why BCS has them #3.

3

98.8

(84/86 – 0/86)

12

Virginia Tech (Bye)

6

98.6

(68/69 – 1/69)

13

Colorado (Lost to Iowa St)

96.7

(58/60 – 2/60)

14

West Virginia (Beat Cincinnati)

12

96.4

(53/55 – 2/55)

15

Fresno State (Beat Boise St)

17

93.8

(30/32 – 2/32)

16

Louisville (Beat Rutgers)

18

93.6

(51/55 – 3/55)

17

Texas Tech (Lost to Oklahoma St)

19

97.1

(50/52 – 1/52)

18

TCU (Beat UNLV)

14

97.6

(59/62 – 0/62)

19

Wisconsin (Lost to Iowa)

20

96.7

(57/61 – 0/61)

20

Minnesota (Beat Michigan St)

21

91.8

(50/55 – 4/55)

21

Arizona St (Lost to UCLA)

89.1

(41/46 – 5/46)

22

Georgia Tech (Lost to Virginia)

93.0

(51/57 – 2/57)

23

Auburn (Beat Georgia)

13

92.6

(50/54 – 4/54)

24

Georgia (Lost to Auburn)

16

88.2

(44/51 – 5/51)

25

Oklahoma (Beat Texas A&M)

25

93.6

(50/55 – 2/55)

Notre Dame(9) at 32. Michigan(15) at 27. Florida(22) at 29. South Carolina(23) at 38. Boston College(24) at 37.

Last place: New Mexico State.

12 Responses to NCAA 1-A Week 12 Beatpaths Top 25

  1. Pat says:

    I don’t really think it’s insane that Alabama is ranked over LSU. There’s a really common argument with teams that each have 1 loss, where Team A beats Team B, and Team A loses to some crappy Team C. They both have one loss, but Team B’s one loss is to a better team than Team A’s.

    Which is essentially exactly what the ranking is saying, of course.

  2. Thok says:

    As for Stanford’s rating, I assume your ranking is ignoring their loss to UC Davis, which artificially inflates their rating.

  3. ThunderThumbs says:

    Well, nothing is ignored per se. I don’t know the ins and outs of every team’s schedule, but every win/loss is accounted for somehow. I’d guess that Stanford lost to UC Davis, then created a beatloop to them (by beating a team who had beaten UC Davis), and then possibly finally formed an alternative beatpath to UC Davis on top of that.

  4. ThunderThumbs says:

    Yes, that has to be what happened – remember the only beatloop teams that are counted are the ones that aren’t already in one of the team’s beatpaths (either to the team or from the team).

  5. ThunderThumbs says:

    Hmm… looks like this ranking finds it less of a surprise that Georgia Tech beat Miami than the BCS did. I have #22 beating #11, BCS had unranked beating #3.

  6. Thok says:

    UC Davis is Division 1AA-you very well might not have included their game with Stanford. Hence why I mentioned it.

  7. ThunderThumbs says:

    Ahhh. Thanks, you’re right. I include it when I calculate all the college teams, but not division 1-A only. Wow, can’t believe Stanford lost that game.

    Right now I only include games where Div 1A plays Div 1A. It’s a shortcut, but that’s one of the potential flaws.

    I’m going to work on something where I consider all games, but then shrink out the irrelevant teams – that’ll be a good improvement, plus, it’ll enable the system to focus on certain divisions, too.

  8. ThunderThumbs says:

    It looks like under normal circumstances, the beatloop that would be taken out would be:

    Portland St=>UC Davis=>Stanford=>Oregon St=>Portland St

  9. ThunderThumbs says:

    Ok, after a bunch of hacking, I’ve figured out a way to incorporate ALL the college football games (even intradivision) and then only examine the div 1-a teams.

    It turns out it actually has a fairly large impact. This approach is much better than ignoring all teams outside div 1-a entirely.

    For the rankings, the first eight are the same, but then it starts changing. Stanford’s actually ranked #54, not #9.

    Others:
    Fresno St at #9 instead of #15 (BCS #17).
    #10 – #14 are the same.
    #15 – #17 are bumped up a notch due to Stanford disappearing.
    Florida jumps from #29 to #18 (BCS #22).
    Oklahoma from #25 to #19 (BCS #25).
    Wisconsin and Minnesota fall from #19, #20 to #20, #21 (same as BCS)
    Arizona St falls out of my rankings. (BCS unranked)
    Kansas and Maryland are at #22 and #23 (BCS unranked)
    Georgia Tech falls from #22 to #24 (BCS unranked)
    Auburn at #25 from #23 (BCS #13)

    Others: Georgia (16) at 26, South Carolina (23) at 28, Michigan (15) at 29, Notre Dame (9) at 31, Boston College (24) at 32.

  10. Pat says:

    How are you handling the Div IAA teams? Are they included in the loop calculations, and then ignored on the output?

    I’m a bit surprised at Notre Dame falling that low. They’re usually ranked low by most algorithms (they… really haven’t beaten that many people, and the loss to Michigan State puts them in a class of much weaker teams) but falling to 31 from 9?

  11. ThunderThumbs says:

    It was BCS who had them #9. The last algorithm had them at 32, the new one at 31.

  12. ThunderThumbs says:

    And yeah – all the beatwin and beatloop numbers are based off of all of all divisions. And it took a while to come to the insight of how to do this simply, but it works out pretty well. I just go through each team, and if it’s not a team that I want to end up displaying in the graph, then I just connect its direct beatlosses directly to its direct beatwins and delete the team. By running the same program multiple times, one for each conference, I should be able to come up with one graph per conference. I’m just not ready to do that yet because I want to try and increase performance first. I don’t relish spending five hours coming up with the conference graphs. 🙂

    I didn’t want to just draw the conference graphs directly by focusing on conference records because it brought up the possible flaw of Team A beating Team B in conference, but of there being a B->C->A beatpath where C is out of conference. The conference graph would show A ahead of B, when really they’d be in a beatloop.

Leave a Reply

Your email address will not be published. Required fields are marked *