Numbers and eye tests

I’m more than familiar with the old saw “lies, damned lies and statistics”, but you can still count me in the group that finds advanced stats useful to evaluate college football teams.  In most years, anyway.  The problem for me now is how much can you rely on stats in a year that’s proven to be the mother of all anomalies?

Here’s how F+, which is the mashup of Bill Connelly’s SP+ and Brian Fremeau’s FEI metrics, sees the nation’s top teams.  (Per the site, “FEI ratings (FEI) represent the per-possession scoring advantages a team would be expected to have on a neutral field against an average opponent, based on opponent-adjusted possession efficiency data. SP+ ratings (SP+) are tempo- and opponent-adjusted measures of college football efficiency based on play-by-play and drive data designed to be predictive and forward-facing.”)

Screenshot_2020-12-31 2020 COLLEGE FOOTBALL F+ RATINGS Football Outsiders

Essentially, Cinci and Georgia are seen as fairly equally matched opponents.  Hell, maybe they are.  But color me a little skeptical about the data in a year like this one.  Essentially, I’ve got three concerns with comparing the two:

  1. Sample size is always something of a concern, even in normal seasons.  2020 is anything but a normal season.  Teams played fewer than the usual 12-game schedule, meaning the data points are fewer, too.
  2. Scheduling restrictions make it harder to analyze the data.  Georgia played a conference only schedule.  Cincinnati didn’t play a P5 program.
  3. COVID and opt-outs meant roster size varied over the course of a season even more than usual.  It’s been fun mocking Florida this morning, but let’s face it — the team that played Oklahoma was a different one than the team that played Georgia.

Taken altogether, I’m left with the impression that the value of such stats this year is on the micro level (conference play) more so than the macro level (national play).  To illustrate, here are the final numbers on SEC net yards per play I compiled:

Screenshot_2020-12-31 SEC net ypp, Week 13

As a guide to judging within the conference, it’s a decent road map:  one truly dominant team, another three good teams, each with their own set of flaws, and the rest ranging from merely mediocre to downright awful.  My overall sense is that 2020 was a down year for the conference overall and I think the ypp numbers bear that out.  But are those numbers useful in relating conference strength to other schools and conferences?  It’s hard for me to say, even though I know Bill and Brian do a careful, considered job to make that case.

Ironically enough, I was looking to the bowl games to help make that case, but, at least in the SEC’s situation, between the cancellations and opt-outs, that’s becoming almost impossible to assess.  I suppose there’s a measure of validation one way or the other to be gained from watching the Peach Bowl.  Then again, small sample size, right?



Filed under Stats Geek!

7 responses to “Numbers and eye tests

  1. Smoky Joe Would

    I find it interesting that the SEC has a down year statistically the year it only plays conference games. Maybe that best conference top to bottom narrative isn’t so far fetched after all.

    Either that or cupcakes really pad those stats for everybody everywhere.

    Or the Big Ten is really the ACC, just more conceited.

    Liked by 3 people

    • originaluglydawg

      “Or the Big Ten is really the ACC, just more conceited.”
      Man, that is pure gold! It needs to become a meme or slogan or something!


  2. originaluglydawg

    If Vandy hadn’t hid from the Dawgs, how would those stats change?


  3. uga97

    #’s are available conference play only year to year. I noticed our scoring offense this year that Monken has increased by 10 points per game vs Coley’s Offense Pts/gm. vs SEC only. This with a 4th string QB & Pickens out several games, a poor showing at UK & no Vandy. Good sign.

    Liked by 1 person

  4. kingcmo2000

    It would be helpful if the bowls were approached and played more consistently by the competing teams just to see how good these analytical tools work in a year like this. However, for many valid reasons bowls are not played that way, so I don’t think we’ll be able to take much away from the results.
    From an advanced stat perspective I think 2020 will be a giant shrug emoji. That will be true across all kinds of disciplines.

    “Excluding 2020” will probably be a parenthetical we should get used to.


  5. willypmd

    I would be interested to see what the net yards per play was with JT at QB for the common opponents for UF/Bama/A&M

    I have a feeling the gap would be closer between us and Bama and we’d be a clear 2nd team…