theknightswhosay

Ranking Teams and Quality Wins

In Bowls, College Football, College Football Playoff, Rankings, Rankings Commentary on December 7, 2017 at 6:03 PM

When I updated my ratings, I was really not happy that two teams that don’t even belong in the conversation were fourth and fifth while the team I thought was the deserving #4 was a somewhat distant seventh.

If Alabama were fifth, I would have shrugged it off. Last year, after the Army/Navy game, Washington edged out Penn St. for fourth. I think if Penn St. had played Washington at a neutral site last December, Penn St. would have won. I know Penn St. lost the Rose Bowl, but a team going from playing for a national championship to playing for nothing but a bowl win can sometimes be a bit of a letdown. I still think the Penn St. team that won the Big Ten Championship game would have beaten Washington, and I still think that they had a better resume as well.

Penn St.’s dramatic Big Ten championship over Wisconsin convinced me they belonged in the top 4 last year. (Pictured: TD catch by TE Mike Gesicki)

No ratings system is perfect though, and for the two to be so close that the result of one major game (there were also a couple of FCS results added in) could tip the balance was good enough. Also, I don’t really mind the deciding factor when two teams are close being who has fewer losses.

But for the arguments I presented over the weekend in Alabama’s favor and to have them that far behind was cause to reevaluate things.

I still think I have a really good formula, but approaching each game neutrally has some shortfalls. So I’m going to have two different computer ratings from now on. I considered “power rating”, but I never know what that means, so I’m just going to call it weighted and unweighted. The new rating will be weighted toward success against the best teams.

With it unweighted, you get the same credit this year for beating Texas and Temple as you do for beating Alabama and having a bye week. Central Florida beat a lot of teams like Texas and Temple but didn’t even play any teams that were nearly as good as Alabama, Ohio St., USC, etc.

This made Wisconsin’s and Central Florida’s 12 wins apiece hard to overcome even though as I pointed out, each only had two wins apiece against the top 40. I could not devise a system I believed in that put Ohio St. ahead of Wisconsin and Central Florida, but by weighting overall strength of schedule and quality wins, I was able to get Alabama ahead of them.

Central Florida-Memphis was a fun game to watch; but being that this was the best team the Knights beat, UCF should not be considered one of the best teams.

When I tried to alter the system to allow for more losses without a high penalty in order to push Central Florida and Wisconsin down, it pushed up teams like USC, Notre Dame, and Auburn instead of Ohio St.

This year, the Big Ten’s problem was depth. Ohio St. only played 7 teams in the top 80 (one out of conference) and Wisconsin played 6. Both Ohio St. and Wisconsin played 10 games against Big Ten opponents, so it should have been higher. By contrast, Alabama played only 8 SEC games and had 9 opponents in the top 80 (Tennessee was the only SEC opponent outside; Florida St. and Fresno St. are both inside the top 80). Oklahoma played 10 games against the top 80 and played 10 conference games (Ohio St. and Tulane are in; Kansas and Baylor are not). Anyway, moving Wisconsin and UCF down a peg for not having very deep schedules is part of the reason Ohio St. fell just slightly below Alabama.

People will see me as a Big Ten detractor, but again, I wanted two Big Ten teams in the playoff last year. Also, in both the weighted and unweighted top 9s from last year, there were two other Big Ten teams. Like this year though, I did think the SEC was the best conference top to bottom.

How do I figure out which games to add weight to? I mentioned the top 80 above, that roughly corresponds with the positive numbers in my unweighted system. So that’s where I drew the line. It so happens to be just low enough to encompass teams with wins over competitors in the major conference (and in one case a playoff team). So along with Syracuse, the low positive-numbered teams also include Cal (Berkeley), which beat Washington St., and Pittsburgh, which beat U. Miami.

So that’s one tier. The next tier includes teams like Ole Miss, Duke, Utah, and Virginia, .500 Power 5 teams. Both Duke and Virginia had good non-conference wins too. It also includes Southern Mississippi, who went 8-4 despite playing two SEC teams in non-conference play.

The next tier starts with teams better than 0.3, which right now is the top 45. This is low enough to include Iowa St., which beat both Oklahoma and TCU. It has some slightly better (than the previous tier) Power 5 teams like Kentucky, UCLA, and Texas A&M as well as Group of 5 teams that competed for titles like South Florida, North Texas, and Fresno St. This tier has a little bit more of an increase in points than the last one.

The penultimate tier is teams better than 0.55. I made it that instead of an even 0.6 because last year it would have only encompassed 18 teams rather than 24. Right now it encompasses 25 teams instead of 23. Numbers 24 and 25 are San Diego St. and Virginia Tech, so this had nothing to do with trying to tilt the playing field. Clemson would have been #1 anyway.

I treated teams better than 0.9 (roughly top 10) a little bit differently. I think whether you win or lose to a top 10 team, you should get a little bit extra consideration. The loss hurts in some parts of the formula, and maybe it shouldn’t hurt as much. So that’s why Auburn and Ohio St. didn’t seem to get the proper credit in the unbalanced formula for playing really good teams out of conference. Also, teams like Wisconsin and Central Florida have less impressive resumes for not beating any top 10 teams.

I’m going to show the final top 10 for both this year and last year. If I had to make a list of teams most likely to compete well in a playoff now and at the end of last year, I’m not sure the membership of either top 10 or top 5 would be any different. Clemson would win obviously, but if I forced Clemson to be higher last year, the formula would move Central Florida higher this year. The whole point was to help teams with good opponents. Clemson and Penn St. were both safely in the top 4 and nearly tied last year, so I’m fine with how it turned out.

Now
1 Clemson 39.717752 (1)
2 Georgia 38.920924 (2)
3 Oklahoma 33.422577 (2)
4 Alabama 32.829833 (7)
5 Ohio St. 32.801820 (6)
6 UCF 32.794786 (5)
7 Auburn 32.580250 (10)
8 USC 32.413394 (8)
9 Wisconsin 32.266683 (4)
10 Notre Dame 32.227370 (9)

December 11, 2016
1 Alabama 43.268628 (1)
2 Ohio St. 37.306985 (3)
3 Penn St. 35.458147 (4)
4 Clemson 35.410836 (2)
5 Wisconsin 30.033484 (9)
6 Washington 28.799632 (5)
7 Colorado 27.670309 (11)
8 Michigan 26.841181 (7)
9 Florida St. 26.795726 (10)
10 Oklahoma 25.479138 (8)
(Western Michigan fell from 6th to 15th)

If you couldn’t tell, the number at the end is where the teams fell in my unweighted ratings.

If you’re interested in the full lists, they are as follows:
2016 Pre-Bowl
2016 Final
Current

I think with so many more opponents in flux or not yet attaining enough points (for instance, only 4 teams qualified for the second-highest tier in my first list this year), the weighted version may be less useful early on. I plan to start publishing both lists at the usual times (Saturday night or early Sunday morning between about October 1 and December 15 and the night of the national championship) though.

Advertisements

Championship Week Top 25

In Bowls, College Football, College Football Playoff, Rankings, Rankings Commentary on December 3, 2017 at 3:42 PM

More on “Who’s #4”

Last night, I think I made sufficient arguments for Alabama to be chosen over Ohio St. I wanted to elaborate on that a little bit more based on some arguments I’ve heard and considered since.

The point was made that Alabama isn’t like last year’s Ohio St., which was ranked #2 going into championship week. Nonetheless, between the two, Alabama was 5th going into championship week while Ohio St. was 8th. So rankings going in is one of the weaker arguments I’ve seen, but I’m just addressing it because it seemed to get some play in social media.

I mentioned how Alabama is favored by Las Vegas over Ohio St. if the two were to play on a neutral site. Alabama was also the #1 most likely to win the national champion and is #1 in ESPN’s Football Power Index. Also, Alabama is in the top 4 in “strength of record,” which compares success versus a given schedule to how the average top 25 team would do. I agree with the “strength of record” calculations that 11-1 against Alabama’s schedule was harder to do than 11-2 against Ohio St.’s. So those are additional reasons I think it’s much harder to leave Alabama out than two-loss Ohio St.

Nick Saban and his team receive the CFP trophy from the late John Saunders after defeating Clemson in January 2016. The Tide will attempt to win its second such trophy in three years.

This isn’t really covering new ground, but I made a chart that I think might be useful in trying to clarify my position.

Team Rank Last SoS Top 40 %/100
Clemson 1 2 12 6-0* 1.00
Georgia 2 3 11 4-1+ 0.80
Oklahoma 3 4 75 4-1 0.80
Wisconsin 4 1 66 2-1 0.67
UCF 5 7 65 2-0 1.00
Ohio St. 6 11 31 4-2 0.67
Alabama 7 4 41 3-1 0.75
USC 8 12 34 3-2 0.60

*lost to (74) Syracuse
+beat (41) Kentucky

I think with the small number of top-40 games, you can eliminate Wisconsin and Central Florida (UCF). Then you can eliminate USC based on the top-40 percentage to get us to the main decision, which was between Alabama and Ohio St.

But for illustration, I wanted to explain why I think Alabama should be ahead of a team with one fewer loss than they have.

Alabama’s one loss was to a better team than any UCF played, so I think it’s hard to count that loss as a point in UCF’s favor. Also, even if that weren’t the case, being 24 spots higher in strength of schedule would result in another loss for any team this year. Maybe a really dominant team from years past could have played a much tougher schedule without losing another game, but there is a reason UCF is the only undefeated team. No team could withstand a much tougher schedule without losing to SOMEONE.

When you go to Ohio St. vs. Alabama, it’s not as easy to get around the extra loss. Alabama beat two teams, LSU and Mississippi St., that are better than Iowa. That’s not just my opinion; that’s the outcome of a range of objective measures as well as the CFP rankings themselves.

Also, 10 spots in strength of schedule makes is harder to argue that if you make Alabama’s schedule a little tougher that they would have lost another game. Maybe if they had played a top-10 team on the road the same day they played Mississippi St. on the road, it would have mattered. If LSU or Fresno St. or any other opponent were a little better (LSU in some ways played better, but they would have had to be a much better team to get 14 more touchdowns or stop Alabama from scoring 14 of its points), Alabama’s record doesn’t change.

Having a better winning percentage against the top 40 is more relevant to fitness for the CFP playoff as well. It’s roughly the top third of teams. Actually the top third is 43 teams if you round down the fraction, but that’s why I noted that Georgia beat Kentucky. It’s also teams that in general are able to beat the better teams (Syracuse/Clemson was obviously an outlier). If these teams have a bad day against a team not in the top 40, there is a very high chance they win anyway.

Oklahoma played a lot of easy opponents obviously, which is why I have their strength of schedule below that of both Wisconsin and Central Florida; but they made up for it with a very top-heavy schedule. Three wins in the top 15 is hard to do, and none were in doubt for much of the second half. I’m not sure I can explain the loss to Iowa St. very well, but the Cyclones are a good enough team that if you get them on the road they can be dangerous to anyone under the right circumstances. Just ask Mike Gundy, who lost to a much more mediocre Cyclone team to miss out on a chance to play for the national championship with Oklahoma St. in 2011.

I think Clemson’s and Oklahoma’s respective losses put into context Alabama’s loss to Auburn (a much, much better team than either of those losses) and close win over Mississippi St. (a team that at least would be the clear favorite against either Iowa St. or Syracuse regardless of location).

Rankings from 9 to 25

Notre Dame fell short of a New Years Six Bowl, but this win over eventual Pac-12 champion USC helps make them the best of the rest. (Pictured: RB Josh Adams)

It’s a very close call between Notre Dame and Auburn (0.00071), but I do think Notre Dame had a slightly harder schedule. It will be interesting to see what the Irish do against another SEC team that beat Auburn (they lost to Georgia by 1 in September and play LSU on January 1 in Orlando).

Miami is not far behind the Irish. I know they blew out Notre Dame, but they didn’t beat USC by 35 either. The Hurricanes lost to a much worse team (Pitt) before the ACC title game than any teams who beat Notre Dame (Georgia, Stanford, and themselves).

Penn St. didn’t have any great non-conference or even non-divisional opponents, but they played in a pretty tough division and were one point away from a chance to make the playoff.

I mentioned Stanford a moment ago. They were better than Notre Dame and pretty even with USC in recent weeks, but it’s not that easy to overcome three losses especially when one was to San Diego St. (although the Aztecs are now in my top 25). I would have rather seen the Cardinal in a New Years Six bowl than Washington, but no one asked me.

Boise St., Florida Atlantic, and Toledo moved up with wins in conference championship games. I think we could make a chart like I did with the top 8 and explain that they’re not really in the top 20 of toughest teams to beat, but what I focus on is a system that general gets top few teams right. This is the first year in a while that I’m not happy with 1 through 4 although I think 1 through 3 are perfect. Anyway, the point is I’m not going to alter my ratings to make 3- or 4-loss teams higher in this part of the rankings. That would move teams like Notre Dame and Auburn higher, which I don’t want to do.

The rankings after the bowl sort out some of these issues because the major-conference 3- or 4-loss teams generally get better opponents than the minor-conference champions.
Odd for this to happen after such a dramatic week, but there was no turnover at all in my top 25.

rank/team/prev.
9 Notre Dame 8
10 Auburn 5
11 U. Miami 9
12 Penn St. 10
13 Stanford 13
14 Washington 14
15 Boise St. 20
16 Mich. St. 15
17 Fla. Atlantic 25
18 Memphis 16
19 TCU 17
20 Toledo 23
21 Wash. St. 18
22 LSU 19
23 Northwestern 21
24 San Diego St. 22
25 Virginia Tech 24

Full list

What the CFP Playoff Should Look Like

In College Football, College Football Playoff, Rankings, Rankings Commentary on December 2, 2017 at 10:16 PM

Sugar: Clemson vs. ??
Rose: Georgia vs. Oklahoma

Before I get to the discussion, I wanted to start by acknowledging that my formula is not always going to align with which teams are deserving of the playoff. I do think it had the top 4 teams correct going into the week, but I think in a way championship week functions as a preliminary round of the playoffs, and Wisconsin was rightly eliminated.

If the Big Ten had not had a championship game and this were just another week of conference play, maybe Wisconsin would be deserving of being ranked ahead of Ohio St. just like last week I thought Georgia and Alabama deserved to be ahead of Auburn. But when you win a championship game over the other team, that should overcome having one more loss. So had Georgia been undefeated going into the SEC Championship Game and Auburn won, I’d be willing to overlook that extra loss by Auburn.

Let’s not forget last year. Penn St. beat Ohio St. and had one more loss, but they did not play one another for the conference title. So Penn St.’s conference championship did not eliminate the Buckeyes from the national title hunt.

Also, I think it’s fair to say that if there are at least four teams with better schedules and one loss or fewer, you can eliminate teams that do not have schedules in the top 50, which neither Wisconsin nor Central Florida have.

Anyway, if you eliminate Wisconsin and Central Florida, 4 and 5 are Ohio St. and Alabama. Ohio St. is ahead on my list, but I don’t compensate in any way for the fact that Ohio St. got an extra game this week. If you average the rating by playing week, Alabama comes out ahead of Ohio St. instead of one spot behind.

That doesn’t mean Alabama should automatically be considered better, but when it’s a conflict to me that means we need to look further. One reason I didn’t rank Wisconsin #1 on my blog sooner was the fact that they were not first when you divided by number of playing weeks, so that is something that I’ve looked at this year as well as in other years.

As I’ve made clear over the years, I’m no fan of Nick Saban or Alabama despite their being in the SEC. The last time Ohio St. played Alabama, I was for the Buckeyes. So I’m not writing in favor of Alabama out of any personal desire for the Tide’s success.

This article is a good breakdown of all of the statistical arguments in favor of Alabama.

It doesn’t mention that if you want to focus on the top 10 alone, Ohio St. is now 2-1 and Alabama is 0-1; but would you really ignore Alabama’s 11-0 record otherwise? Would you also ignore that Ohio St. lost to a team that isn’t even in the top 25 by 31 points just 4 weeks ago?

Whether it had to do with injuries, the change of location, or the defense just being ready for the way Auburn plays, I don’t think what we saw from Auburn today was the same thing the Tide faced last week.

I hate to be reminded of this, but I think if you look at what Georgia did today, it has some similarities to the adjustments Alabama was able to against LSU in the 2011 season. So I think if you gave Alabama a second chance against Auburn at a neutral location, they would also win.

I didn’t like how Gary Danielson talked about Alabama all night or how he said we shouldn’t look at the SEC Championship game in evaluating Alabama, but Auburn certainly didn’t become inferior to Iowa all of a sudden. I think he was right to say Auburn may have beaten anyone with the way they played at home their last two SEC regular-season games.

If you factor in Iowa, you need to factor in teams that are better than Iowa like LSU and Mississippi St. I don’t care if they’re not “signature wins”, but they’re wins over better teams than the one that dominated Ohio St. four weeks ago.

Another way I can see to rationalize it is to say we should look at wins only. I think if we do that, it’s easy to pick Ohio St. If we pretend Alabama and Ohio St. had the same number of losses, I don’t think that tells you who is likely to be a good team in the playoff, which I think the standard should be when two teams are close enough statistically to get into a discussion like this. The losses are informative about the range of performances that you can expect in the playoff.

In short, I think varying between losing to Auburn by 12 and beating LSU by 14 is better than varying between losing to Iowa by 31 and beating Wisconsin by 6.
I’m not crying for Alabama if they’re left out though. They’ve backed into enough national-championship competitions as it is.