BYU SOS: trends and watering?

[pelagius]

There has been some talk about SOS schedule and thing like watering down the schedule. So let's take a look at it.

Here is a look at BYU historical strength of schedule is end of year MOV computer model rankings of its opponents.

This graph show plots two variables. First, the average year end MOV computer model ranking of BYU's opponents. Trying to capture trends in the "typical opponent" played by BYU. Second, is how many times per year BYU plays a top 40 opponent. Trying to capture trends in the quality of the opponent.



The trends are a little easier to see using a 4 year moving average:



Just looking at the average, the 2006-2009 era has the easiest schedule. The average opponent rank during those 4 years has climbed to over 80. Which is easier than any era faced by Edwards. However, 2006-2009 does better with the equality opponent metric with an average around 3. nt terms of quality opponents the era has been very typical. It is similar to the early 80s.

The schedule now is clearly much easier than the early 90s. BYU played a much tougher schedule during the last few years of the Detmer era and the first couple years after Detmer. This is true for both average opponent and the number of quality opponents.

Is this watering down? Sure. Certainly relative to the Detmer era. However, the biggest effect is showing up in the average. A lot of that has to do with how bad the conference is now outside of BYU, Utah, and TCU.

[Indy Coug]

Two comments:

1. "Watering down" implies there is an active effort to make our schedule easier. Unless you can demonstrate that BYU deliberately tried to schedule UCLA, Washington, Florida State, Oklahoma, etc. during very down years and that they have significant control over how good the remainder of the conference is, you can't make the claim that the schedule has been "watered down".

2. The number of D1-A opponents over the last 25 to 30 years has increased by roughly 20%, so you would need to also adjust the SOS rankings to a percentile rather than a raw SOS value.

[TheBYUGuy]

Two comments:

1. "Watering down" implies there is an active effort to make our schedule easier. Unless you can demonstrate that BYU deliberately tried to schedule UCLA, Washington, Florida State, Oklahoma, etc. during very down years and that they have significant control over how good the remainder of the conference is, you can't make the claim that the schedule has been "watered down".

2. The number of D1-A opponents over the last 25 to 30 years has increased by roughly 20%, so you would need to also adjust the SOS rankings to a percentile rather than a raw SOS value.

I would add that even though the sample size is significantly lower, we should really be looking at OOC schedules since those are the ones that BYU has some control over. Aren't we trying to determine whether BYU has been scheduling down rather than how good conference opponents are? (I'm assuming we're trying to answer the question posed in another thread.)

[pelagius]

Two comments:

1. "Watering down" implies there is an active effort to make our schedule easier. Unless you can demonstrate that BYU deliberately tried to schedule UCLA, Washington, Florida State, Oklahoma, etc. during very down years and that they have significant control over how good the remainder of the conference is, you can't make the claim that the schedule has been "watered down".

The Washington and UCLA examples work But not the OU example and probably not the Florida St examples. For example OU is ranked is the top 10 in the sample for this year (MOV victory computer models love them). Florida St is 42. Which is probably pretty a little lower than their average when scheduled. So the measures I am using aren't really over-penalizing BYU they way you are suggesting.

You're right about the watering down thing ... I don't really mean it. Which why I tried to suggest that a lot of the weakening was coming from the average conference opponent getting worse.

2. The number of D1-A opponents over the last 25 to 30 years has increased by roughly 20%, so you would need to also adjust the SOS rankings to a percentile rather than a raw SOS value.

Kind of ... not really .. there were a 110 in 1984. Top 40 is still a reasonable metric of a quality opponent. It will have an effect on the average though.

[pelagius]

I would add that even though the sample size is significantly lower, we should really be looking at OOC schedules since those are the ones that BYU has some control over. Aren't we trying to determine whether BYU has been scheduling down rather than how good conference opponents are? (I'm assuming we're trying to answer the question posed in another thread.)

I'm not ... I really don't even care about watering down. I thought my response of "Sure" was meant to signal I don't care that much about it but I should have been more explicit.

[Indy Coug]

Kind of ... not really .. there were a 110 in 1984.

Are you sure? What is your source on that? I can recall Utah in the 1980s having the worst or next-to-worst defense in the nation and was ranked in the mid-100s, not the 110s.

You could also take into account there were several D-1 teams in the 1980s (and early 1990s?) that no longer exist, such as Long Beach State, CS-Fullerton, Pacific, etc.

[pelagius]

Are you sure? What is your source on that? I can recall Utah in the 1980s having the worst or next-to-worst defense in the nation and was ranked in the mid-100s, not the 110s.

I am going by Jim Howell's list of teams and years (not the rankings, just the list)

http://www.jhowell.net/cf/cfindex.htm.

In the late 70s even the Ivy league was considered 1-A ... I believe 1-AA was created in the late 70s. You see more teams go 1-AA in the early 80s. I think the bottom is like 106 in 88 or something like that. The big change is in the 1-AA ranks. Those ranks have really swollen. If you include 1-AA in the rankings then you have bigger comparison problems.

You could also take into account there were several D-1 teams in the 1980s (and early 1990s?) that no longer exist, such as Long Beach State, CS-Fullerton, Pacific, etc.

Not sure why one would want do induce survivorship bias?

[pelagius]

This solves most of the potential issues related to changes in number of teams in division 1-A. I just bin everything and form an index.

1, rank <= 20
2, 20 rank <= 40
....
6, rank > 100

You still get basically the same thing althoug the difference aren't as big as before:

moving averages

       year      mrank   mtop40   mtop20  
  1.   1976          .        .        .  
  2.   1977          .        .        .  
  3.   1978          .        .        .  
  4.   1979   4.333333     1.75      .75  
  5.   1980   4.020408     1.75       .5  
  6.   1981   3.882353      2.5       .5  
  7.   1982       3.88     2.75      .75  
  8.   1983        3.8     2.75      .75  
  9.   1984        3.7     2.75      .75  
 10.   1985   3.490196     2.75      1.5  
 11.   1986   3.461539     2.75      1.5  
 12.   1987   3.509434        3     1.25  
 13.   1988   3.377358     3.75     1.75  
 14.   1989   3.423077      3.5     1.25  
 15.   1990   3.384615     3.75     1.25  
 16.   1991   3.173077     4.25     2.25  
 17.   1992   3.230769     4.75        2  
 18.   1993   3.098039        5      2.5  
 19.   1994       3.08        5      2.5  
 20.   1995     3.3125     3.75      1.5  
 21.   1996       3.44     3.25     1.75  
 22.   1997   3.673469     2.75      1.5  
 23.   1998       3.88      2.5      1.5  
 24.   1999    3.72549      3.5     1.75  
 25.   2000   3.583333      3.5      1.5  
 26.   2001   3.764706      3.5        1  
 27.   2002       3.88     2.75       .5  
 28.   2003       3.84      2.5      .75  
 29.   2004   3.755102     2.75     1.25  
 30.   2005   3.468085     3.25        2  
 31.   2006   3.395833     4.25     2.25  
 32.   2007   3.612245     4.25     1.75  
 33.   2008   3.901961     3.75     1.75  
 34.   2009   4.098039     3.25      1.5

You still get that relative to the early 90s, the typical team is about 20 ranks below. This doesn't mean I don't think BYU has performed well in the last 4 years. They certainly have met my expectations. But BYU is playing a lot of bad teams.

[jay santos]

I know Rhondo had a strategy to pretty much max out the SOS. Hale/Bateman/Skousen/Crowton dumbed that down a little by dropping some key games that were scheduled and adding crap games. Bronco/Holmoe/Eyring started in the direction of dumbing it down even more. 9 of the first 10 teams (or something like that) Bronco scheduled were non-BCS. He was vocal about following the "Hawaii model". I think he changed his philosophy about a year or two ago.

So I agree the data doesn't really show there's been a dumbing down of the schedule, but the philosophies were known and there was a dumbing down in the philosophy (even if the execution or actual outcome doesn't show it).

My scheduling whining is mostly due to our crappy conference situation and there's not much anyone can do to satisfy me.

[pelagius]

So I agree the data doesn't really show there's been a dumbing down of the schedule, but the philosophies were known and there was a dumbing down in the philosophy (even if the execution or actual outcome doesn't show it).

Well the drop in the average opponent quality is being driven being by crappiness of the typical conference team. However, the number of quality opponent in conference I bet is at on all time high. So quality from out of conference must have dropped. Certainly some of it could be bad luck in scheduling but that shouldn't be driving a 4 year moving average.

[pelagius]

For those that were wondering here is non-conference games (dropping bowl games). Just non conference games:

4-year moving average for each one

       year      mrank   mtop40   mtop20  
  4.   1979   4.176471      .75       .5  
  5.   1980   3.833333      .75       .5  
  6.   1981      3.625      .75       .5  
  7.   1982          4      .75       .5  
  8.   1983   3.857143       .5       .5  
  9.   1984   3.769231       .5       .5  
 10.   1985     3.3125        1      .75  
 11.   1986   3.176471     1.25      .75  
 12.   1987   3.176471      1.5       .5  
 13.   1988   3.117647     1.75      .75  
 14.   1989     3.1875     1.75       .5  
 15.   1990      3.125     1.75       .5  
 16.   1991     2.8125        2     1.25  
 17.   1992   2.529412     2.75     1.25  
 18.   1993   2.235294        3     1.75  
 19.   1994        2.2     2.75      1.5  
 20.   1995   2.571429     2.25      .75  
 21.   1996          3      1.5      .75  
 22.   1997   3.083333     1.25      .75  
 23.   1998          3     1.25      .75  
 24.   1999          3      1.5      .75  
 25.   2000        2.5        2      .75  
 26.   2001      3.375      1.5      .25  
 27.   2002   3.722222     1.25      .25  
 28.   2003        3.4      1.5      .75  
 29.   2004       3.35      1.5        1  
 30.   2005   2.736842     2.25      1.5  
 31.   2006        2.5     2.75      1.5  
 32.   2007       2.75      2.5        1  
 33.   2008   3.882353     1.75       .5  
 34.   2009      4.125     1.25      .25

The difference here is striking relative to the early 90s. BYU71 really does have something to complain about (BYU averaged 1.25 top 40 opponents per year between 2006-2009). Sure some/substantial portion of it could be bad luck in terms of teams having bad years when they happen to play BYU (certainly 2001 and 2008, but they moving average should overcome this to some degree). However there looks like a real shift after 2005. 2004 and 1991 are tied for the best year in non-conference opponent quaility. 2005 is still good and then it shifts.

Here are yearly results which, of course, are much noiser

       year      arank   top40   top20  
  1.   1976        5.5       0       0  
  2.   1977        5.5       0       0  
  3.   1978        2.6       1       1  
  4.   1979        3.5       2       1  
  5.   1980          4       0       0  
  6.   1981        5.5       0       0  
  7.   1982   3.666667       1       1  
  8.   1983          3       1       1  
  9.   1984       3.75       0       0  
 10.   1985          3       2       1  
 11.   1986          3       2       1  
 12.   1987          3       2       0  
 13.   1988        3.5       1       1  
 14.   1989       3.25       2       0  
 15.   1990       2.75       2       1  
 16.   1991       1.75       3       3  
 17.   1992        2.4       4       1  
 18.   1993          2       3       2  
 19.   1994          3       1       0  
 20.   1995   3.333333       1       0  
 21.   1996       3.75       1       1  
 22.   1997          2       2       2  
 23.   1998   2.666667       1       0  
 24.   1999       3.25       2       0  
 25.   2000          2       3       1  
 26.   2001          5       0       0  
 27.   2002        4.2       0       0  
 28.   2003   2.333333       3       2  
 29.   2004       1.75       3       2  
 30.   2005        2.5       3       2  
 31.   2006        3.5       2       0  
 32.   2007       3.25       2       0  
 33.   2008        5.8       0       0  
 34.   2009   3.333333       1       1

[pelagius]

I take back what I said early ... Most of it is not driven by the have-nots in conference getting worse on average. It is a bit worse, the big difference is the out of conference results:

Conference for year moving average:

       year      mrank   mtop40   mtop20  
  1.   1976          .        .        .  
  2.   1977          .        .        .  
  3.   1978          .        .        .  
  4.   1979    4.62069       .5        0  
  5.   1980   4.275862       .5        0  
  6.   1981     4.1875        1        0  
  7.   1982   4.030303     1.25        0  
  8.   1983    4.03125     1.25        0  
  9.   1984   3.878788      1.5        0  
 10.   1985   3.806452        1      .25  
 11.   1986    3.83871      .75      .25  
 12.   1987    3.90625      .75      .25  
 13.   1988       3.75        1       .5  
 14.   1989    3.78125      .75      .25  
 15.   1990       3.75        1      .25  
 16.   1991    3.59375     1.25      .25  
 17.   1992   3.870968        1        0  
 18.   1993   3.833333        1        0  
 19.   1994   3.677419      1.5       .5  
 20.   1995   3.774194        1       .5  
 21.   1996   3.764706     1.25       .5  
 22.   1997   3.971429     1.25       .5  
 23.   1998   4.371428      .75      .25  
 24.   1999   4.264706     1.25      .25  
 25.   2000    4.21875        1      .25  
 26.   2001     4.1875     1.25      .25  
 27.   2002   4.133333        1        0  
 28.   2003   4.206897      .75        0  
 29.   2004   4.107143        1      .25  
 30.   2005   3.964286        1       .5  
 31.   2006          4     1.25      .75  
 32.   2007    4.16129     1.25      .75  
 33.   2008   4.129032     1.25        1  
 34.   2009     4.3125     1.25        1