The Linked Letters After Dark: Week 11 Playoff Edition
Do the college football playoff committee's decisions in this week's rankings make sense?
The college football playoff committee’s new rankings are out, so it’s time to examine the decisions made by the committee. I made the case for BYU being a playoff team in my article a few hours ago, but the committee’s current rankings would leave them without a playoff spot. I’d like to analyze both this and a few of the committee’s other decisions that don’t make sense to me.
ESPN posted an article discussing five teams that David Hale believes have cause to be angry over their ranking. Much of this article revolves around comparisons involving Teams A and B, where Team A is inevitably the one that has cause for anger and Team B is a higher ranked team. In my opinion, this style of criticism involves picking metrics that are favorable to Team A while paying less attention to those that support ranking Team B higher. The current rankings are really an update to last week’s rankings, so my question is whether the changes from last week to this week make sense.
BYU, Virginia, and Louisville all lost games on Saturday and proceeded to drop five spots. Iowa’s close loss at Oregon only dropped them one spot. Ole Miss also dropped one spot, but that’s partly due to Texas Tech moving up two spots following a quality win over BYU. Are these reasonable updates to the rankings?
Ole Miss and Texas Tech
Does it make sense for Texas Tech to move ahead of Ole Miss this week? Texas Tech has a marginally better strength of record and is ranked 14 spots higher in the predictive rankings. It’s hardly crazy to place Texas Tech ahead of Ole Miss, but can that change be justified on this past weekend’s results?
Similar to my calculations for strength of schedule and strength of record, let’s choose a benchmark team with a predictive rating approximately 1.5 standard deviations above the FBS mean. This week, I’ll give that hypothetical team a predictive rating of 72 to keep the numbers a bit simpler, placing the team between #12 BYU and #13 Iowa. The home advantage is 2.15 points, and I’ll include that in the calculations. I’ll refer to this hypothetical team as Team X, and I’ll predict their expected margin of victory against an opponent using this formula: Expected Margin = 72 +/- 2.15 (advantage or disadvantage) - Opponent Rating. The strength of result uses this formula: Strength of Result = Real Margin - Expected Margin For Team X, such that positive numbers are a stronger result.
For Texas Tech against BYU, the expected margin for Team X is 72 + 2.15 - 72.29 = 1.86 points. The real margin was 22 points, so the strength of result is 22 - 1.86 = 20.14 points.
For Ole Miss against The Citadel, the expected margin for Team X is 72 + 2.15 - 19.22 = 54.93 points. The real margin was 49 points, so the strength of result is 49 - 54.93 = -5.93 points.
Overall, Texas Tech’s result was 26.07 points better than Ole Miss’ result, which is no small difference. I have no problem with the committee moving the Red Raiders ahead of the Rebels this week.
Miami and Georgia Tech
Vanderbilt, Miami, Georgia Tech, USC, and Michigan all moved up due to Virginia and Louisville each falling five spots in the rankings. In the process, Miami also passed Georgia Tech. Miami has a better strength of record and predictive rating than Georgia Tech, but does that change make sense the past weekend’s results?
Georgia Tech didn’t play, so their strength of result is effectively zero for the week. Miami won 38-10 over Syracuse. Team X’s expected margin is 72 + 2.15 - 47.03 = 27.12 points. Therefore, Miami’s strength of result is 28 - 27.12 = 0.88 points. Because Georgia Tech didn’t play, this is effectively saying that Miami deserves to move ahead of them on the basis of a 0.88 point strength of result. Although I have no problem with the Hurricanes being ranked ahead of the Yellow Jackets, this seems like a strange time to make that change in the rankings. If anything, this looks more like the committee correcting a perceived mistake that Miami was ranked too low instead of Miami earning it on this past weekend’s result.
BYU, Virginia, Louisville, and Iowa
With the exception of Iowa, the three other teams currently ranked that lost games last weekend all dropped five spots in the rankings. In past years, it’s almost seemed like there’s a standard penalty of four or five spots in the rankings for losing a game. This week, that standard penalty appears to be five spots, with Iowa being the lone exception having lost a very close game against Oregon.
For BYU at Texas Tech, the expected margin for Team X is 72 - 2.15 - 79.44 = -9.59 points. The real margin was 29-7 in favor of Texas Tech, so the strength of result is -22 - (-9.59) = -12.41 points.
For Wake Forest at Virginia, the expected margin for Team X is 72 + 2.15 - 55.68 = 18.47 points. The real margin was 16-9 with Wake Forest winning, so the strength of result is -7 - 18.47 = -25.47 points.
For California at Louisville, Team X’s expected margin is 72 + 2.15 - 46.73 = 27.42 points. The real margin was a 29-26 win for California, meaning that the strength of result is -3 - 27.42 = -30.42 points.
And then for Oregon at Iowa, Team X’s expected margin is 72 + 2.15 - 83.28 = -9.13 points. The actual margin an 18-16 Oregon win, meaning that the strength of result is -2 - (-9.13) = 7.13 points.
I understand that teams generally drop in the ratings after a loss, but it’s actually hard to justify penalizing Iowa any spots for this result. Still, a one spot decline is the smallest penalty the committee could give if they’re going to lower a team’s ranking for a loss, and that’s what the committee did to Iowa.
But does it make sense for BYU to fall the same number of spots as Virginia and Louisville? Although BYU’s loss was by a larger margin, it’s actually a stronger result due to the difficulty of playing a road game at Texas Tech. But BYU was also ranked higher, so it probably wouldn’t be justified to drop them as much as Virginia or Louisville even if the strength of result was identical for all three teams.
For the most part, ratings are normally distributed in the FBS. It means there are a few really good teams at the top with their ratings spaced farther apart, then a lot of mediocre teams with ratings closer together, and then a few bad teams with ratings spaced farther apart again. There’s a 7.61 point difference from #1 Indiana to #4 Oregon in my predictive ratings. However, a predictive rating 7.61 points below #11 Miami goes all the way down to between #26 Missouri and #27 Pittsburgh. Bill Connelly applies a seven point penalty per loss in his backward looking ratings derived from SP+, which is reasonable. But that seven point penalty will also cost a team more positions if they’re toward the middle of the FBS ratings than if they’re near the top. Even if BYU’s loss was just as bad as Virginia’s and Louisville’s losses, they probably shouldn’t lose as many spots just because BYU is closer to the tail of the distribution.
That said, BYU’s loss clearly isn’t as bad as Virginia’s loss. And Virginia’s loss isn’t quite as bad as Louisville’s loss. If Virginia should lose five spots and Louisville’s loss was worse, then it might make sense to flip Iowa and Louisville in the rankings. As a result, Iowa doesn’t get penalized for their close loss to Oregon, which had a positive strength of result. And Louisville loses six spots instead of five due to the particularly poor strength of result.
Using my logic for Virginia and Louisville, it means that falling one spot in the rankings is worth about -5 points in the strength of result. Instead of BYU dropping five spots, it then makes more sense for them to drop just two or three spots. This puts them either at #9 or #10 behind either Oregon or Notre Dame, respectively. In either case, BYU is still in the playoff. But because BYU is farther out on the tail of the distribution, I’d favor dropping them only two spots instead of three. That puts BYU at #9.
Overall, I believe that BYU should only fall to #9, Virginia would fall five spots to #19, Iowa would stay at #20 instead of falling a spot, and Louisville would fall six spots to #21.
Group of 5 Teams
I lamented that last week the committee didn’t include any Group of 5 teams in their rankings. That has changed this week with South Florida entering the rankings at #24. However, there’s a very good case for North Texas, James Madison, and Tulane to be in the top 25. It’s an improvement over last week, but I still believe the Group of 5 is underrepresented. My playoff ratings have both South Florida and North Texas in the top 25, and James Madison is also in the top 25 in my strength of record calculations. I have no problem with Cincinnati entering the top 25, and I actually put them at #21 in my subjective rankings along with the three aforementioned Group of 5 teams. The real issue is the complete omission of the Group of 5 in last week’s playoff rankings and the committee having limited opportunity to correct it this week with only Missouri and Washington falling from the top 25.
There’s quite a bit of variability in different computer ratings, so the committee deserved some flexibility in their initial ratings. I didn’t really agree with how low they ranked Miami or how high they ranked Utah. But aside from the Group of 5 omission, I believe the committee generally deserved the benefit of the doubt for an initial ranking that’s subjective, and where there’s always going to be a case that some teams are ranked in the wrong place. But that initial rating is the starting point, and each week is then an update from the previous week instead of a completely new set of rankings. Therefore, I expect that the updates will then be done in a reasonably consistent and logical manner, but it seems like BYU paid a particularly harsh penalty for their loss in Lubbock. And that’s my biggest complaint this week other than the Group of 5 still being underrepresented.
I’ll post more content later this week, including updated NFL ratings tomorrow. Thanks for reading!