Comparing Preseason AP, Coaches, and SP+ Rankings

Submitted by Blue@LSU on August 16th, 2023 at 4:29 PM

The final SP+ rankings for the preseason dropped today, so it seemed like it would be interesting to see how they compare to one another. Here’s the graph (sorry, it's pretty congested):

Florida and Kentucky make an appearance in SP+, but not AP or the Coaches Poll.

It can also fun to look at each team’s offensive and defensive SP+ rankings. Another graph:

In case anyone was wondering, here is the graph SP+ rankings for all B1G opponents and Michigan’s OOC opponents.

NittanyFan

August 16th, 2023 at 4:42 PM ^

The visual presentation is very well done.  :-)  +1

Maybe I'm not as up to date on Maryland as I should be --- but SP+ projects they have a (by a non-insignificant margin) better defense than offense?

CityOfKlompton

August 16th, 2023 at 7:34 PM ^

Simply by the numbers, Maryland's defense was a little better than their offense last season, but I must stress that's by the numbers.

Looking at TeamRankings, they really didn't stand out anywhere, but they were mostly solid in some high level metrics. Pair that with a decent time of possession ranking, and you can see why Maryland's defense ranks a little higher than expected in an efficiency ranking that, at this point, is largely based on last season's performance.

Obviously, that doesn't tell the whole story, but you can see how it makes sense (as well as the flaw in SP+ rankings.)

bronxblue

August 16th, 2023 at 9:03 PM ^

Yeah, per SP+ they ended last year 33rd on defense and 54th on offense, and while I didn't think their offense was super-efficient against the better teams on the schedule they certainly didn't seem particularly stout against any good teams they saw.  

AFAICT the one thing they seemed pretty good at was not giving up big plays; they were a top-25 team in giving up 20+ plays. on defense, and so I guess having teams walk down the field in small-ish steps is more efficient than giving up huge busts.  Now, is that reproducible given their turnover on the roster as well as, at least to my untrained eye, nothing particularly innovative on defense?  Probably not, and the year before (2021) they were a bit below-average nationally stopping the big play.  So we'll see.  

I also do think SP+ has some weird quirks built into it and so when I see teams that went 8-5 with mediocre top-line numbers but shiny-ish advanced stats I sort of assume the numbers are finding a pattern or a proof of concept that might not be there.  Maryland has athletes on defense and there's a base level of competence there but that wasn't a top-35 defense out there, much like how Duke was a top-40 defense last year per SP+ mostly because they had a +16 TO margin.

NittanyFan

August 17th, 2023 at 1:01 AM ^

Good stuff ...... +1.

My "Maryland experience" last year was basically pieces of 4 games.  (1,2) Wisconsin & PSU: The opposition got up multiple scores in the 1st half, the weather was ugly, the opposition realized "it's November and this Maryland offense can't move the ball", so they parked the bus in the 2nd half.  (3,4) Michigan & OSU: The final score-line looked respectable, yardage difference as well, but I always had the sense the opposition offense would easily score/have success when they really had to.

So, thinking it over more: I can see why an algorithm might like the MD defense, even though we didn't while watching.

blueblooded14

August 16th, 2023 at 4:54 PM ^

This might look foolish in a few months but I think OSU's offense will take a step back this season (note - take a step back from #1). They have so much talent but I'm a bit skeptical about how their qb play will be, especially early in the season behind new tackles. I'm not dumb enough to doubt that OSU will have a high-functioning offense by November, but not the best in the nation.

 

ChiBlueBoy

August 16th, 2023 at 4:55 PM ^

I've always had the impression that SP+, particularly early in the year, wasn't very reliable. Am I imagining that, or does anyone have a sense of how accurate it has proven to be retrospectively (e.g., comparing SP+ preseason rankings vs. game outcomes or final rankings)?

NittanyFan

August 16th, 2023 at 4:59 PM ^

It's not --- nothing really is great in August --- but IMO it's still a useful pre-seasonal directional tool.

Teams lose 30%+ of their production yearly, new coaches, new schemes, team chemistry changes, all of that stuff is hard to precisely quantify.  Several games of empirical data make the metrics considerably better.

ak47

August 16th, 2023 at 7:57 PM ^

The pre season poll has historically been better at predicting the top 25 at the end of the year than the early week polls. Pre season expectations are by and large built on pretty solid facts and then people vastly overreact to the small sample size that is the first 2-3 weeks of the season 

NittanyFan

August 17th, 2023 at 12:48 AM ^

Your first sentence isn't wrong.  But I'd argue that's not because the "pre-season expectations are built on pretty solid facts", but rather because "in the early-season, voters get distracted easily by a shiny undefeated record against a less-than schedule."

AP Poll last year: Kentucky (!) once rose to #7 last year, so did Oklahoma State, Ole Miss & UCLA to #9, NC State to #10, BYU got to #12 after beating #9 Baylor, Syracuse got to #14, Kansas to #19, and even James Madison was once #25.

What all above have in common: they reached their peak while undefeated.

I definitely think SP+ (and FEI) gets less distracted by zeroes at the end of a record.

Ranking my personal "order-of-trust": it would be SP+ late season, SP+ mid-season, polls late season, SP+ early-season, polls mid-season, SP+ preseason, polls preseason, polls early-season.

Amazinblu

August 16th, 2023 at 4:57 PM ^

Is it possible that Northwestern's offense will actually be lower than 125?   And, continuing with Northwestern - 30th for Defense this season seems optimistic as well.

MeanJoe07

August 16th, 2023 at 4:58 PM ^

I like this visual.  It pleases me a great deal. Like that feeling when you spray scalding hot water on your eczema rash. That second or two where it just feels so good.  I have a lot of rashes.  That's what you get when you hang out in trees all day rubbing against bark. 

Tex_Ind_Blue

August 16th, 2023 at 5:03 PM ^

Michigan's opponents are horrible in offense. But I am not sold on their defense either. That's not the case for the Pre-season top 25, they are better on offense (15ish) than defense (10ish). 

 

Thanks for plotting these. Puts things in a new perspective.  

CarrIsMyHomeboy

August 16th, 2023 at 8:45 PM ^

Delightful!

I really do find it to be the best among a differentially flawed set of available indicators of CFB prowess. So, my interest is both:

  1. genuine about the transition through the first month as SP+ sheds its preseason muck, and
  2. extended well beyond that because SP+ comparisons have an "above average" track record of predicting wins (and even something as modest as a 55% predictive record is a smashing success by comparison to CFB's other non/advanced stats)