Evaluating the College Football Playoff Selection Committee's Week 13 Rankings
How do the computer ratings compare to the selection committee's rankings, which conference is strongest, and how can schedule strength be measured?
Let's begin with a detailed look at the selection committee's new rankings, which were released on Tuesday. This article uses data for games through November 23, even though several games have been played since then. The biggest difference in the college football playoff selection committee's rankings compared to my ratings is the inclusion of Tennessee over BYU. Let's scrutinize these teams in some more detail.
Tennessee vs. BYU
Let's get this out of the way: Tennessee is the better team. They have a higher predictive rating, and it's not that close. If they played BYU at a neutral site, Tennessee would be favored by 8.09 points according to my predictive ratings and have an estimated 71.51% chance of winning. But if this was about selecting the strongest teams, the SEC probably gets five of the top 11 playoff spots, including Ole Miss and Alabama. South Carolina could easily make that six teams. The college football playoff, like most postseason tournaments, includes the most deserving teams instead of just the best teams.
Tennessee and BYU are both 9-2, and BYU is ahead in my strength of record ratings. But let’s take a close look at both team’s schedules and compare them. My schedule strength rating is just an average of the predictive ratings for each team's opponents. Prior to this week's games, Tennessee was ranked #56 by that metric. I suspect Jeff Sagarin's ratings do something similar, where the Volunteers were ranked #57. But ESPN's FPI had Tennessee's schedule at #22. By comparison, my ratings, Sagarin, and FPI ranked BYU's schedule at #30, #21, and #39, respectively. FPI's schedule strength estimates the schedule's difficulty for an average top 25 team, meaning that it's a relative strength of schedule. My strength of schedule metric is absolute, in that it doesn't depend on which team plays that schedule.
Tennessee played three particularly weak opponents: at home against Chattanooga (predictive rating 26.26, would be ranked between #121 and #122), Kent State (8.11, #134), and UTEP (19.21, #128). Tennessee would be favored by several touchdowns in each of these games whereas they were only favored by just under four touchdowns in their neutral site game against North Carolina State (40.65, #76). But does the rating difference between Kent State and North Carolina State really matter in terms of wins and losses? For Tennessee, the answer is they probably don’t matter very much. A team of Tennessee's quality (predictive rating of 66.32, #7) almost certainly wins all four games easily, which they did. The distinction between Kent State and North Carolina State matters a lot if you're Florida State (37.47, #89) but not so much for Tennessee. Perhaps this isn't the most useful way to compare schedules.
BYU (58.23, #20) played a lot more games that were expected to be competitive, but their predictive rating is also lower, and those games wouldn't appear quite as competitive if Tennessee played the same schedule. For Tennessee’s schedule to date, the predictive ratings estimate that Tennessee should win 8.71 games whereas BYU would have 7.49 wins. Against BYU's schedule to date, Tennessee would be expected to win 8.89 games versus 7.39 wins for BYU. These really aren't large differences in schedule strength, and both teams are 9-2 against their respective schedules. Tennessee’s schedule was slightly tougher for them than BYU’s schedule, but it’s not really enough to distinguish the two teams. I don't see a big difference in strength of record, and closely scrutinizing the results of individual games somewhat defeats the point of the holistic approach in the strength of record. At least in terms of schedule strength and results, there’s just not that much of a difference between Tennessee and BYU.
If I'm subjectively deciding which teams get in the playoffs instead of just reporting my strength of record ratings, I'd select Tennessee over BYU. The schedule strength and strength of record are actually very close between BYU and Tennessee, but there's a significant difference in their predictive ratings. This isn't just an indicator of a team's future performance but also a measure of which teams are actually the best teams, and Tennessee clearly wins there. But the difference is not that obvious just by comparing strength of record and schedule strength, and certainly not enough to justify the playoff committee's decision to rank Tennessee at #8 and BYU at #19.
Comparing the Playoff Committee Rankings and Computer Ratings
As with last week, I'm comparing both the strength of record and overall ratings using the same approach. The actual rating is the team’s rating in either my overall or strength of record ratings. The effective rating is what the team’s rating would be using the selection committee’s ranking. For example, if the section committee ranks a team #7, then the effective rating is whatever the rating is for the #7 team. The difference is just the effective rating minus the actual rating. A negative difference means that the committee has ranked a team lower than my rating system, but a positive difference means the committee’s rankings are higher than my system.
Overall Ratings Comparison
Rank Difference Team Actual Effective
1 -4.21 Alabama 66.47 62.26
2 -3.47 Colorado 59.50 56.03
3 -3.09 Ole Miss 64.43 61.33
4 -1.99 Kansas State 59.56 57.57
5 -1.87 Georgia 67.69 65.82
6 -1.84 BYU 61.33 59.50
7 -1.79 Texas 70.16 68.37
8 -1.67 Louisville 57.67 56.00
9 -1.56 LSU 57.57 56.00
10 -1.37 Iowa State 60.93 59.56
11 -1.33 South Carolina 62.26 60.93
12 -0.78 Notre Dame 68.37 67.59
13 0.00 Ohio State 69.35 69.35
14 0.00 Arizona State 60.76 60.76
15 0.04 Indiana 63.94 63.97
16 0.13 Texas A&M 58.63 58.76
17 0.46 SMU 63.97 64.43
18 0.79 Tennessee 63.96 64.75
19 1.34 Tulane 58.76 60.11
20 1.71 Miami 64.75 66.47
Rank Difference Team Actual Effective
21 1.87 Penn State 65.82 67.69
22 2.57 Oregon 67.59 70.16
23 2.61 Missouri 56.03 58.63
24 3.83 Clemson 60.11 63.94
25 4.20 UNLV 53.72 57.92
26 4.54 Illinois 53.13 57.67
27 6.05 Boise State 57.92 63.96 Strength of Record Ratings Comparison
Rank Difference Team Actual Effective
1 -4.44 Colorado 60.72 56.28
2 -4.05 Kansas State 60.44 56.39
3 -4.03 Arizona State 64.96 60.93
4 -3.99 BYU 64.44 60.44
5 -2.86 Iowa State 63.42 60.56
6 -1.75 LSU 57.96 56.21
7 -1.60 Georgia 67.65 66.05
8 -1.50 SMU 66.46 64.96
9 -0.96 Texas 69.08 68.12
10 -0.47 Penn State 68.12 67.65
11 -0.07 Syracuse 56.28 56.21
12 0.00 Oregon 72.12 72.12
13 0.00 Alabama 62.43 62.43
14 0.00 Texas A&M 59.20 59.20
15 0.39 South Carolina 60.56 60.96
16 0.57 Indiana 63.87 64.44
17 0.88 Notre Dame 66.05 66.93
18 0.96 Miami 65.50 66.46
19 1.02 Missouri 57.67 58.69
20 1.45 Illinois 56.21 57.67
Rank Difference Team Actual Effective
21 2.15 Ohio State 66.93 69.08
22 2.28 UNLV 55.68 57.96
23 2.46 Clemson 60.96 63.42
24 2.92 Ole Miss 58.69 61.61
25 2.94 Boise State 60.93 63.87
26 3.89 Tennessee 61.61 65.50
27 4.33 Tulane 56.39 60.72 Alternate FPI-Like Strength of Record Ratings
I'd also like to look at one more metric, which is similar to what I used in my comparison between BYU and Tennessee. FPI strength of record is based on comparing a team's actual record against how an average top 25 team would fare against the same schedule. One of the components that gets averaged into my strength of record ratings is similar, but it uses a benchmark of the mean predictive rating of FBS plus 1.5 standard deviations. This week, that hypothetical team would have a rating of 63.89, which would be just below South Carolina (63.95, #10). It’s not exactly an average of the top 25 teams, but it’s close. The strength of record here is just a team's winning percentage minus the winning percentage expected for a 63.89 rating team against the same schedule. I haven’t calibrated it to the same scale as the predictive ratings like I usually do, but it’s good enough to rank the teams. Here are the alternate strength of record ratings for the top 30 teams:
Strength of Record Against Mean FBS + 1.5 Sigma
Rank Rating Team
1 0.180 Georgia
2 0.179 Oregon
3 0.126 Ohio State
4 0.119 Penn State
5 0.119 Texas
6 0.085 Miami
7 0.073 SMU
8 0.070 Indiana
9 0.062 Notre Dame
10 0.058 Tennessee
11 0.045 BYU
12 0.037 Boise State
13 0.025 Alabama
14 0.014 Arizona State
15 0.013 South Carolina
16 0.012 Army
17 0.003 Clemson
18 0.003 Iowa State
19 -0.018 Texas A&M
20 -0.019 Ole Miss
Rank Rating Team
21 -0.024 Missouri
22 -0.039 LSU
23 -0.042 Kansas State
24 -0.058 Tulane
25 -0.065 Illinois
26 -0.075 UNLV
27 -0.079 Colorado
28 -0.084 Florida
29 -0.087 Oklahoma
30 -0.089 Louisiana This, of course, does not take into account Georgia's wild 8 OT game against Georgia Tech last night or any of the other games since last weekend. Boise State gets a boost from #16 in my normal strength of record ratings to #12 here. Both BYU and Tennessee are still in the playoffs, but at the expense of Arizona State. Alabama and South Carolina aren't far behind, either. Army is ranked #28 in my normal strength of record ratings but jumps to #16 here.
In the rating comparisons, a negative number means that my ranking of a team is above the committee's ranking. Where are the biggest differences between my ratings and the college football playoff rankings?
Georgia Bulldogs
Georgia was at the top of the list last week for teams underrated by the selection committee, but they moved up in the committee's ratings this week. It remains to be seen how much their rating changes after narrowly getting past Georgia Tech with a win, but they're probably in the playoff regardless of what happens in the SEC championship game.
Big 12 Teams
An ESPN article Tuesday night speculated on the possibility that both Boise State and Tulane might get into the playoff at the expense of any Big 12 team. That scenario isn't going to happen since Tulane lost to Memphis. Still, it seems like the Big 12 is systematically underrated by the selection committee. Colorado, BYU, Kansas State, and Iowa State are all higher in my ratings than the committee's rankings. I went through a lengthy discussion comparing BYU and Tennessee, and I couldn’t justify the large difference between the teams in the selection committee’s rankings. Although Arizona State’s predictive rating isn’t that strong, they also have compiled an impressive record, and arguably are underrated by the committee as well.
The Big 12's top teams aren't as strong as the best teams in the other power four conferences, but the conference also doesn't have any particularly weak teams. The Big Ten and ACC have Purdue (28.33, #117) and Florida State (37.47, #89), which are ranked below the weakest Big 12 team (Arizona; 40.14, #78) in the predictive ratings. There's a good argument that the Big 12 is somewhat underrated, though the scenario of the conference getting shut out no longer seems possible. Would the selection committee rank the Big 12 teams higher if there were a couple of top 10 teams in the conference like the SEC, ACC, and Big Ten? It’s possible that this works against the committee’s ranking of Big 12 teams.
LSU Tigers and Texas Longhorns
Although I don't believe the SEC is anywhere close to being systematically underrated like the Big 12, the committee did Texas and LSU no favors here. Their ratings come at the expense of teams like Miami and Tennessee.
Tennessee Volunteers
I believe that Tennessee probably doesn't belong quite as low as #14, like my strength of record ratings, but I also don't believe they should be ranked as high as #8. My alternative strength of record ratings, which are an approximation of FPI strength of record, have Tennessee at #10, which seems a bit more reasonable.
Miami Hurricanes
I'm not entirely convinced that Miami is overrated, but both the normal strength of record and the overall ratings have them at #8. Like SMU, Miami only has one loss, and that loss was against Georgia Tech. Because I'm posting this early Saturday morning, I should also point out that Georgia Tech might well rise in the next ratings after nearly winning at Georgia, meaning that Miami's record may look even more impressive.
Clemson Tigers
Clemson is ranked at #15 in my normal strength of record ratings and #17 in my overall ratings. The alternative strength of record ratings also have the Tigers at #17. Although I made a strong argument that Tennessee's schedule is better than it initially appears, that argument doesn't work for Clemson. The Tigers' strongest opponents were Georgia and Louisville, and they lost easily to both teams. None of my ratings have them anywhere close to #12, which is where the selection committee ranked Clemson. If they win against South Carolina, it would probably boost their rating a bit, but it's not clear that this should make them a playoff team. Unlike Miami's loss to Georgia Tech or SMU's loss to BYU, neither of Clemson's losses were especially close.
UNLV Rebels and Tulane Green Wave
Before losing to Memphis, Tulane had achieved a good predictive rating despite a weak schedule. UNLV's schedule is even a bit weaker. Regardless of which strength of record rating I use, Tulane is at #24. UNLV is a bit lower, at #26 in the alternate strength of record ratings, and #27 in the normal ratings. Both seem to have been overrated by the committee.
Boise State Broncos
Unlike UNLV and Tulane, Boise State deserves their own discussion, and they’re not clearly overrated. The normal strength of record ratings have the Broncos at #16, but they do rise to #12 when taking into account the difficulty that a better team would have against their schedule. Neither rating puts the Broncos all the way up at #11, but they're not clearly overrated by several spots like UNLV and Tulane. They will likely play UNLV for the Mountain West championship, and if they win that game, their only loss will have been by three points to Oregon. It’s hard to consider the Broncos overrated even if some of my ratings suggest they are.
Conference Ratings
Let’s aggregate the predictive ratings and try to calculate which conferences are strongest top to bottom. The SEC has a lot of teams in the top 10 or 20 according to the predictive ratings. The Big Ten has a few highly ranked teams, but the ratings seem to drop off considerably after the top four teams. The ACC and especially the Big 12 don’t have as many teams near the top of the ratings. But does that mean the SEC and Big Ten are stronger than the ACC and Big 12?
There are a few ways to rank the quality of conferences, and the simplest is just to take the average predictive ratings of all the teams in the conference. The problem is that a simple mean can be skewed by having one very strong or very weak team in the conference. The median rating avoids this problem, but it's still not necessarily representative of the entire conference, because the median is only determined directly by the rating of one or two teams. Jeff Sagarin uses what he calls the central mean, giving more weight to teams ranked toward the middle of the conference while still using the ratings of all the teams in the calculation. This is the right idea, because it attempts to solve the issues with just using the median or a simple mean, and is a better measure of which conferences are the strongest overall.
I've taken a slightly different approach, instead hypothetically assuming every team in a conference played a neutral site game against every other FBS team or every other team in the entirety of my ratings. I use the predictive ratings to calculate the expected winning percentage for the entire conference in those games. The rankings can be a bit different depending on whether the comparison is just against the FBS or against nearly all of college football, but it should be less prone to being skewed in the same way a simple mean is.
The "Win%" column is a conference’s expected winning percentage if each team in the conference played every team in the FBS at a neutral site. The "All CFB" column is the same, except for every team ranked by my system, across all divisions. The "Mean" column is the mean predictive rating for the teams in the conference, and the "Median" column is the same but for the median rating.
Expected Conference Winning Percentage against FBS Opponents
Rank Win% Conference All CFB Mean Median
1 .7441 SEC .9315 (1) 59.03 (1) 57.47 (1)
2 .6442 Big 12 .9060 (2) 51.84 (2) 52.19 (2)
3 .6228 Big Ten .8982 (3) 51.09 (3) 49.76 (3)
4 .5978 ACC .8927 (4) 49.32 (4) 48.39 (4)
5 .4956 FBS Independents .8432 (5) 44.06 (5) 40.09 (5)
6 .4186 Pac-12 .8381 (6) 38.70 (6) 38.70 (6)
7 .3965 Sun Belt .8265 (7) 37.11 (7) 38.25 (7)
8 .3668 American Athletic .8088 (9) 35.07 (9) 34.26 (8)
9 .3597 Mountain West .8140 (8) 35.16 (8) 32.64 (9)
10 .2976 Mid-American .7794 (10) 30.41 (10) 29.56 (10)
11 .2518 Conference USA .7594 (11) 27.48 (11) 29.33 (11)The SEC is the clear winner here, but there's not really that much of a gap between the Big 12, Big Ten, and ACC. The SEC and Big Ten have their mean ratings skewed upward compared to the median because of having several teams rated highly in the predictive ratings. There's a big gap between the SEC and the other power four conferences. Then there's another gap that's even a bit larger to the group of five conferences. The most interesting point here besides the relative strength of the conferences is that having a several highly ranked teams doesn't necessarily represent the overall quality of the conference. The SEC has several highly ranked teams, and the conference is ranked at the top here. But the balance of the Big 12, which lacks highly ranked teams but has many good teams ranked in close proximity to each other, is actually #2 on the list. This doesn’t necessarily mean the Big 12 should get more teams in the playoff, but it’s a sign that the Big 12 isn’t as weak as the team ratings would suggest. For that matter, the ACC isn’t that weak, either.
Measuring Schedule Competitiveness
Just like schedule strength can be measured, and I’ve said plenty about that already, it's also possible to measure schedule competitiveness. The first step is to decide what makes a game competitive, or how big the margin of victory has to be before it's considered a blowout. I do this by fitting a statistical distribution to the margin of victory for every game that's been played this season and use that to determine a benchmark for what is a competitive game. This season, that works out to around 17.14 points, or roughly two touchdowns and a field goal. For each game, I calculate the probability that the margin of victory is likely to be within this threshold. Even for two teams that are exactly evenly matched, there's still a possibility that the game will still turn into a blowout, but this scenario also gives the maximum probability that any game will be competitive. The competitiveness rating is calibrated so that 0% is a game that has no possibility of being competitive, and 100% means that the two teams are as evenly matched as possible.
I also calculate a dominance rating, which goes from -100% to 100%. For every game, there's some chance of it being competitive, but also a chance that each team will dominate the other. For both teams, I calculate the probability of them winning in a blowout and the probability of a blowout loss. If there are equal chances of a blowout win and a blowout loss, the dominance rating is 0%. If a team is much more likely to lose in a blowout than win in that fashion, the rating becomes negative. If a team is likely to dominate their opponent, the rating is positive.
The competitive and dominance ratings aren't measuring what happened in each of a team's games, but what would be expected based on the team's rating and their opponents' ratings. The ratings for each game are averaged over a team's entire schedule to calculate the overall competitive and dominance ratings. I also calculate a team's expected winning percentage given their rating and the rating of all of their opponents. The "Rating" column is the competitiveness rating. The "Predictive" column is the predictive rating for each team. The "Dominance" column ranks whether a team tended to have ratings well above (positive) or well below (negative) their opponents. Finally, "ExpectedWin%" is a team’s expected winning percentage based on their predictive rating and the predictive ratings of their opponents.
Competitive Game Ratings
Competitive game threshold: 17.14 points
Rank Rating Team Predictive Dominance ExpectedWin%
1 90.40% Old Dominion 41.83 (74) -8.66% (88) .494 (80)
2 88.45% Coastal Carolina 34.06 (99) -15.01% (92) .419 (93)
3 88.19% West Virginia 49.64 (48) 4.94% (79) .471 (86)
4 87.81% Virginia Tech 55.53 (29) 32.49% (51) .652 (38)
5 87.58% New Mexico 30.87 (107) -24.52% (93) .411 (97)
6 87.35% Kansas 57.36 (25) 30.06% (54) .597 (52)
7 86.96% Texas Tech 48.36 (53) 10.17% (75) .507 (77)
8 86.79% Syracuse 47.22 (56) 20.54% (65) .551 (63)
9 85.96% Cincinnati 49.99 (45) 5.79% (77) .493 (81)
10 85.71% Colorado 58.27 (19) 45.90% (34) .708 (24)
11 85.48% USC 56.69 (27) 41.47% (41) .671 (33)
12 85.43% Florida International 30.70 (108) -2.14% (83) .535 (68)
13 85.04% North Carolina 47.32 (55) 29.21% (56) .587 (53)
14 84.93% Arizona State 56.57 (28) 42.76% (38) .674 (31)
15 84.71% Toledo 38.10 (85) 34.17% (47) .611 (48)
16 84.50% App State 34.68 (96) -32.22% (102) .375 (102)
17 84.27% Georgia Southern 42.02 (72) 0.56% (82) .523 (72)
18 83.94% Utah 48.32 (54) 11.17% (72) .509 (76)
19 83.12% Liberty 32.71 (101) 46.20% (33) .686 (29)
20 82.85% UCF 52.83 (35) 25.60% (57) .549 (65)
Rank Rating Team Predictive Dominance ExpectedWin%
21 82.80% Sam Houston 35.22 (95) 17.16% (67) .577 (55)
22 82.55% San José State 35.29 (94) 4.21% (80) .507 (78)
23 82.39% Minnesota 53.95 (33) 41.92% (40) .644 (42)
24 81.94% UCLA 45.52 (60) -42.65% (111) .310 (117)
25 81.80% Nevada 34.41 (98) -24.95% (94) .447 (91)
26 81.68% Houston 41.86 (73) -45.52% (115) .299 (118)
27 81.67% Oklahoma State 43.75 (67) -27.62% (99) .358 (108)
28 81.50% TCU 51.55 (38) 33.17% (50) .609 (49)
29 81.20% Fresno State 36.99 (91) 16.20% (68) .564 (60)
30 81.15% LSU 57.17 (26) 22.46% (61) .543 (67)
31 80.95% Boston College 49.74 (47) 23.98% (58) .569 (57)
32 80.88% Florida 57.52 (23) 8.85% (76) .509 (75)
33 80.38% Baylor 55.39 (30) 45.02% (35) .651 (39)
34 80.32% Rutgers 45.01 (63) 10.76% (73) .480 (84)
35 80.20% Air Force 29.38 (112) -32.55% (103) .372 (104)
36 80.12% Oregon State 32.14 (104) -27.22% (98) .415 (95)
37 79.71% Buffalo 29.47 (111) -12.70% (89) .460 (88)
38 79.45% California 48.39 (52) 39.97% (44) .647 (41)
39 79.39% Central Michigan 25.40 (122) -36.16% (107) .364 (107)
40 79.03% Kansas State 58.67 (17) 49.82% (28) .677 (30)
Rank Rating Team Predictive Dominance ExpectedWin%
41 78.81% North Texas 35.59 (92) -6.35% (85) .479 (85)
42 78.79% Michigan 50.60 (41) 10.37% (74) .520 (73)
43 78.67% Eastern Michigan 26.97 (120) -7.33% (86) .459 (89)
44 78.45% East Carolina 35.39 (93) 32.45% (52) .606 (51)
45 78.07% NC State 40.65 (76) -14.96% (91) .460 (87)
46 77.88% Louisiana Tech 27.96 (118) -5.68% (84) .490 (83)
47 77.61% Virginia 43.75 (66) -28.17% (100) .406 (98)
48 77.41% Georgia State 31.76 (106) -52.56% (118) .294 (120)
49 77.20% Pittsburgh 50.50 (42) 30.04% (55) .568 (58)
50 77.19% Wisconsin 49.48 (49) 5.31% (78) .490 (82)
51 76.90% Iowa State 58.44 (18) 58.55% (23) .729 (20)
52 76.82% Maryland 42.74 (69) -25.04% (95) .417 (94)
53 76.57% Wyoming 27.84 (119) -54.62% (119) .312 (116)
54 76.53% Arizona 40.14 (78) -41.96% (110) .341 (111)
55 76.52% Troy 32.00 (105) -31.42% (101) .371 (105)
56 76.31% Duke 46.79 (57) 23.88% (59) .567 (59)
57 76.00% BYU 58.23 (20) 53.45% (26) .672 (32)
58 75.98% Washington State 45.25 (61) 51.08% (27) .691 (27)
59 75.84% Auburn 52.31 (36) 34.52% (45) .584 (54)
60 75.76% Washington 48.64 (51) 13.37% (70) .527 (69)
Rank Rating Team Predictive Dominance ExpectedWin%
61 75.59% South Alabama 42.33 (71) 43.75% (36) .656 (37)
62 75.42% UConn 40.09 (79) 47.64% (30) .662 (34)
63 73.96% Colorado State 34.47 (97) 2.64% (81) .525 (71)
64 73.88% Georgia Tech 50.37 (43) 20.24% (66) .545 (66)
65 73.84% Texas A&M 58.07 (21) 49.81% (29) .647 (40)
66 73.72% Rice 32.34 (103) -26.90% (97) .391 (100)
67 73.72% Kennesaw State 18.64 (129) -65.60% (126) .225 (129)
68 73.53% Memphis 42.75 (68) 62.92% (20) .744 (19)
69 73.20% Vanderbilt 53.07 (34) 20.79% (64) .504 (79)
70 72.97% Oklahoma 57.43 (24) 22.33% (62) .513 (74)
71 72.42% Louisville 60.35 (15) 55.06% (25) .687 (28)
72 72.10% Northern Illinois 39.01 (83) 30.94% (53) .614 (47)
73 72.09% Jacksonville State 37.75 (87) 47.16% (31) .698 (26)
74 71.86% UL Monroe 32.35 (102) -34.30% (106) .392 (99)
75 71.81% Ohio 39.73 (81) 46.66% (32) .634 (44)
76 71.79% Miami (OH) 40.49 (77) 34.35% (46) .628 (46)
77 71.74% Arkansas State 28.39 (116) -43.21% (112) .378 (101)
78 71.49% Florida Atlantic 22.91 (125) -48.08% (117) .297 (119)
79 71.36% Illinois 50.04 (44) 34.10% (48) .575 (56)
80 71.26% San Diego State 26.66 (121) -43.96% (113) .350 (110)
Rank Rating Team Predictive Dominance ExpectedWin%
81 71.25% UTSA 33.12 (100) 13.29% (71) .550 (64)
82 70.74% Wake Forest 37.85 (86) -39.96% (108) .331 (113)
83 70.14% South Carolina 63.95 (10) 62.44% (22) .715 (23)
84 70.00% Utah State 30.57 (109) -34.17% (105) .413 (96)
85 69.56% Missouri 54.39 (32) 40.68% (42) .607 (50)
86 69.51% Iowa 57.53 (22) 58.55% (24) .718 (21)
87 69.39% Arkansas 51.52 (39) 23.74% (60) .526 (70)
88 69.36% Texas State 46.20 (59) 66.63% (18) .749 (18)
89 69.15% New Mexico State 18.16 (130) -70.71% (128) .217 (130)
90 68.86% Hawai'i 28.74 (113) -26.11% (96) .374 (103)
91 68.65% Nebraska 49.37 (50) 21.87% (63) .552 (62)
92 68.31% Ball State 23.52 (124) -63.28% (124) .278 (123)
93 67.98% UTEP 19.21 (128) -67.29% (127) .266 (126)
94 67.83% Northwestern 39.24 (82) -44.07% (114) .331 (112)
95 67.77% Louisiana 45.07 (62) 62.45% (21) .717 (22)
96 67.45% James Madison 46.50 (58) 70.36% (17) .761 (16)
97 67.25% Western Kentucky 37.06 (90) 33.47% (49) .631 (45)
98 66.98% Georgia 67.73 (6) 64.37% (19) .699 (25)
99 66.09% UNLV 51.75 (37) 71.30% (16) .763 (15)
100 65.84% Marshall 44.71 (64) 42.49% (39) .658 (35)
Rank Rating Team Predictive Dominance ExpectedWin%
101 65.54% Middle Tennessee 17.44 (132) -71.91% (130) .245 (127)
102 65.51% Western Michigan 29.65 (110) -14.55% (90) .449 (90)
103 65.33% Charlotte 28.51 (114) -62.28% (123) .282 (122)
104 65.15% Navy 44.54 (65) 43.23% (37) .657 (36)
105 64.45% Bowling Green 40.88 (75) 40.15% (43) .643 (43)
106 63.29% SMU 61.48 (13) 77.07% (13) .807 (11)
107 62.78% Penn State 63.51 (11) 75.71% (14) .785 (14)
108 62.76% UAB 28.41 (115) -46.90% (116) .326 (114)
109 62.52% Kentucky 50.72 (40) -7.91% (87) .435 (92)
110 62.20% Stanford 37.72 (88) -58.63% (121) .274 (124)
111 62.11% Alabama 70.50 (4) 78.32% (12) .809 (10)
112 61.87% Clemson 59.26 (16) 73.83% (15) .757 (17)
113 61.43% Florida State 37.47 (89) -59.46% (122) .267 (125)
114 59.64% Temple 21.91 (126) -76.05% (131) .233 (128)
115 59.01% Mississippi State 42.44 (70) -41.61% (109) .351 (109)
116 58.96% Akron 23.57 (123) -58.55% (120) .314 (115)
117 58.73% Michigan State 40.05 (80) -32.78% (104) .368 (106)
118 58.03% South Florida 38.89 (84) 13.64% (69) .556 (61)
119 57.39% Oregon 63.06 (12) 80.96% (10) .812 (9)
120 55.50% Massachusetts 21.41 (127) -63.66% (125) .288 (121)
Rank Rating Team Predictive Dominance ExpectedWin%
121 54.93% Miami 64.01 (8) 84.00% (7) .825 (7)
122 54.82% Ole Miss 70.17 (5) 84.53% (6) .834 (6)
123 54.52% Boise State 54.90 (31) 80.52% (11) .790 (13)
124 54.40% Tennessee 66.32 (7) 82.40% (9) .791 (12)
125 53.35% Tulsa 15.69 (133) -71.56% (129) .214 (131)
126 48.38% Southern Miss 17.72 (131) -88.24% (133) .150 (133)
127 47.52% Texas 71.23 (2) 89.56% (3) .874 (2)
128 46.54% Tulane 61.13 (14) 89.28% (4) .851 (4)
129 45.22% Indiana 64.01 (9) 88.23% (5) .840 (5)
130 44.26% Purdue 28.33 (117) -82.06% (132) .200 (132)
131 44.02% Army 49.85 (46) 83.73% (8) .815 (8)
132 42.40% Ohio State 71.77 (1) 91.60% (2) .870 (3)
133 38.48% Kent State 8.11 (134) -91.83% (134) .147 (134)
134 36.30% Notre Dame 70.69 (3) 94.18% (1) .903 (1)
I'm still figuring out what to do with these ratings, and one possibility is another way to evaluate a team's strength of record. Instead of relying on benchmarks like the rating of the average top 25 team or the average FBS team in calculating strength of record, a different approach might involve a team's actual winning percentage, expected winning percentage, and predictive rating.
It's possible to say that one predictive rating system is objectively better or worse than another. The accuracy of predictive ratings can be measured by comparing the predictions to the results of games. It's also possible to objectively compare one set of subjective rankings with another, and I do something fairly close to that when I comment on the selection committee's rankings. But there's really no objective way to say that one set of subjective rankings is better than another, so my goal isn't really to match the selection committee's rankings or the polls. Instead, I hope to create some tools that are useful to evaluate teams, share the data, and then let you decide for yourself which teams are best.
An example of where this might be significant is Notre Dame’s schedule strength. It’s not entirely Notre Dame’s fault, but they’ve played a particularly weak schedule by some metrics. I don’t think anyone would have expected Florida State to collapse the way they did this season, but games like that haven’t helped. Their average opponent rating is 43.58, which is good enough for #58. However, their schedule is last in competitiveness, and has been significantly less competitive than other teams like Ole Miss and Alabama that have similar predictive ratings. Despite being #3 in the predictive ratings and only having a single loss, Notre Dame is #7 in the strength of record ratings, and these metrics provide some justification for this. They simply haven’t been in that many competitive games, and haven’t had too many opportunities to lose games. The competitiveness, dominance, and expected winning percentage ratings probably all need some additional context to be useful, but they’re tools that can help evaluate a team’s previous games. I’ll probably use some of these ratings and add in a new experimental strength of record rating in my next set of ratings on Sunday or Monday.
What's Next?
I'll have another update on the rankings after all of week 14's games are finished. If the Georgia-Georgia Tech game is a preview of what's to come, it will be a fun Saturday. Remember that there's still a lot of crazy things that can happen, so teams that are two or three spots out of the playoff certainly have a path to get in. I'm working on some other tools to evaluate teams and a few other projects that I hope to share very soon. Enjoy the rivalry games this weekend!
If you're interested in more content like this, please click the "Subscribe" button below and share this article on social media. I develop my own code to generate these ratings and predictions, of course relying on data from others and open source libraries, but none of my content is or will ever be created with generative AI. Your help in supporting more quality content like this is greatly appreciated. This article relies on data from CollegeFootballData.com, which contains a wealth of information about college football, makes it freely available, and shares it in formats that are easily accessible. I have also included data from Jeff Sagarin’s college football rankings and ESPN’s FPI.