Ray vs the World

"Welcome back, my friends
To the show that never ends
We're so glad you could attend
Come inside! Come inside!" ~ELP

 Once again, we enter the Twilight Zone of voter induced hysteria. No, I don't mean the last POTUS election. I mean AP, RJ Young, and Massey's Index!

Let's take a look at the AP, shall we?

Ray vs AP

Now, you Sooner fans should not be upset about what I'm about to do to your team, ok? This goes for MSU and Oregon fans, too. This is straight up math and facts.


AP
RRDiff
1GeorgiaSEC21
2CincinnatiAmerican42
3AlabamaSEC63
4OklahomaBig 12139
5Ohio StateBig Ten116
6MichiganBig Ten3-3
7OregonPac-121811
8Michigan StateBig Ten2618
9IowaBig Ten8-1
10MississippiSEC9-1
11Notre DameIndependent1-10
12KentuckySEC2513
13Wake ForestACC7-6
14Texas A&MSEC239
15Oklahoma StateBig 125-10
16BaylorBig 12193
17PittsburghACC203
18AuburnSEC2810
19Southern MethodistAmerican4930
20Penn StateBig Ten10-10
21San Diego StateMountain West243
22Iowa StateBig 12308
23Texas-San AntonioC-USA3411
24Coastal CarolinaSun Belt7450
25Brigham YoungIndependent14-11

I feel the need to point out that the AP voters are a joke. Let's start with Oklahoma:

  • Oklahoma's OW% is .380 (19-31). 
  • Their total margin of victory (not including FCS WCU) is 61 points in 7 games (8.7 pts/gm).
    • Take out TCU and it's 40 points in 6 games (6.7 pts/gm).
    • They had to 
      • hold off Tulane & Nebraska, 
      • come back vs WVa, 
      • hold off KState, 
      • come back vs Texas & KSU.
  • Their OOW% is .713, which means that the teams they've beaten have been beaten by everyone else, too. Self-evident, but I wanted to spell it out.
  • OU's SoS is .491, good for 109th of 130.
    • For reference, UTSA is also 8-0, their OW% is .531 and their OOW% is .542; their SoS is .534 for 74th of 130

This is the team the AP has 4th. So why is OU 4 while UTSA is 23? Pre. Season. Bias. OU started at 2. UTSA didn't. And while UTSA certainly faced smaller schools and probably deserves the lower rank (I have them at 34, as you can see), no way does OU deserve to be 4.

What about Oregon?

  • OW% .500, OOW% .538, SoS .513 for 95th of 130
  • Their OSU win accounts for opponent wins to one loss, so without OSU 
    • the OW% is .432, 
    • OOW% .537, 
    • SoS .467 for 120th or 130

In other words, Oregon is riding a single victory right now, as most of their others are against sub-par teams.

Michigan State?

  • OW% .395, OOW% .679, SoS .490 for 113th of 130.
  • The team they struggled to put away 20-15 lost to Ohio State 54-7. At home. 

But the voters of the AP have MSU only 3 spots behind OSU. 

So let's look at OSU, who the AP has at 5. 

  • OW% .500, OOW% .586, SoS .529 for 81st of 130
  • How about if they beat Oregon? 
    •  OW% .480, OOW% .607, SoS .522 for 85th of 130
    • Only moves them up to 7 for me

There is literally no case you can make that voters "know better." They clearly do not. I get that the AP is trying to 1) stay relevant in the Bill James era, and 2) give us what they think are "what if" ranks. Even those aren't very good mainly because they are self-fulfilling. And how about that clear SEC bias?

  • Alabama drops from 1 to 5 after losing to a team the AP thought did not deserve a top 25 ranking
  • Ohio State drops from 3 to 9 after losing to a team the AP deemed worthy of being ranked 12
  • More relevant, Iowa dropped from 2 to 11 after losing to an unworthy team
  • Penn State dropped from 7 to 20 after losing to another of the unworthy

In the weeks that followed

  • Alabama dropped to 5, then went to 4 then 3
  • OSU dropped to 9, then went 10, 11, 7, 6, 5, 5
  • Iowa dropped to 11 and jumped to 9 this week

But hey, no bias, right?

Mean5.56
Variance177.34
SD13.32


SEC5.83
Big Ten2.00
Pac-1211.00
Big 122.50
ACC-1.50
American16.00
Mountain West3.00

SEC, Pac12, and AAC are all heavily over-ranked. The mean difference is only 5.5, but that SD!

Massey's Composite

I'm a huge fan of Ken Massey. What I do here is purely for my enjoyment, as I think a composite system is a spectacular way to look at teams who almost never play each other. The only thing I might change would be to exclude systems that seek to emulate voters. Shouldn't be hard to find those: just perform a basic variance check against the AP like I did above and eliminate any that are too close!


Massey's
RRDiff
1GeorgiaSEC21
2MichiganBig Ten31
3AlabamaSEC63
4CincinnatiAmerican40
5Ohio StateBig Ten116
6OklahomaBig 12137
7IowaBig Ten81
8Notre DameIndependent1-7
9MississippiSEC90
10Oklahoma StateBig 125-5
11Wake ForestACC7-4
12Michigan StateBig Ten2614
13Iowa StateBig 123017
14KentuckySEC2511
15BaylorBig 12194
16PittsburghACC204
17Penn StateBig Ten10-7
18Texas A&MSEC235
19OregonPac-1218-1
20AuburnSEC288
21WisconsinBig Ten12-9
22UTSA
C-USA3513
23San Diego StateMtn West241
24TexasBig 123410
25VirginiaACC15-10

I look at these and think that we're not so far off from a truly excellent system.

Mean2.52
Variance52.34
SD7.23


SEC4.67
Big Ten1.00
Pac-12-1.00
Big 126.60
ACC-3.33
American0.00
Mountain West1.00

The mean and SD are half of what they are with just the AP (which is a component of Massey's system). Now it's the SEC and Big12 over-ranked, and the ACC under-ranked.

Transitive Property

Had another guy on Facebook tell me that my algorithm sucks because I have ND over Cincy and they beat ND. 

In a world of perfect competition (which your Econ teacher should have explained to you is a fantasy), Cincy would be ranked above the Irish. It's difficult for me to have that conversation with the uninitiated, as they simply do not understand that who you beat matters, and that one great win does not offset beating a bunch of inferior teams (yes, I'm looking at you, Sooners - and you still do not have a great win).

Let's take a look and break it down:

Notre DameCincinnati
FSU0.4154MIAMI (OH)0.5000
TOLEDO0.5538INDIANA0.2077
PURDUE0.8692ND0.9923
WISC0.9077TEMPLE0.3308
VATECH0.5231UCF0.7077
USC0.4538NAVY0.3846
Sum3.7231Sum3.1231
Avg0.6205Avg0.5205

Both teams have 6 wins against FBS schools. Cincy has a win vs FCS Murray St that we do not count.

ND has beaten 6 teams with .6205 average 1st level points, while the Manbearpigs have beaten 6 teams with an average of .5205 1st level points including Notre Dame. It's not about the one game, it's about the body of work. Isn't that what the talking heads preach right up until they say stupid things like "beat x ranked teams" and "eye test?" It's not that complicated, but you have to understand the method and then agree that all teams should be treated equally. How often do you see or hear about SEC or B1G bias? I mean, outside of here. 

All. The. Time.

So here's a proven way to eliminate that, and the results we get are legitimate no matter what your gut tells you because this is not a predictor! My system is not telling you that ND beats Cincy if they play again. It is telling you that the Irish have a better body of work even with the loss.

I guess what you really have to accept (and we can see how well this is going with the whole vaccine thing) is that what you "know" ain't necessarily so!

Thanks for reading!

Ray

No comments:

Post a Comment

Latest Rankings

Latest Rankings

How Ray’s Rankings are achieved: My system is based on the Ohio State High School Football playoff calculator (and I'm sure other state...