Rankings Roundup after Week 9: CFP, AP, SP+ 

Submitted by Blue@LSU on November 1st, 2023 at 1:02 PM

I usually try to post this on Monday, but I wanted to wait for the CFP rankings to come out last night. With the CFP rankings on Tuesday nights, I’ll probably be posting this on Wednesday from now on. So anyway, here are the numbers after week 9.

COMPARING CFP, AP, AND SP+ (RANKINGS)

Astute observers will recognize that I replaced the Coaches Poll with the CFP. I didn’t want the graphs to be too cluttered and, well, the Coaches Poll sucks anyway. 

The CFP committee has decided to buck the trend and put the Buckeyes at #1. IMO they haven’t really passed the eye test to be the #1 team in the country, but I’m fine with it. It’ll all work itself out in the end. 

There were, however, two people that were up in arms about the CFP rankings. It wasn’t about OSU being #1 or Georgia being snubbed at the #2 position. No, their beef was about something else entirely. 

How dare the CFP not consider the integrity of the game!

SP+ OFFENSE & DEFENSE (RANKINGS)

Focusing on the SP+ …

6 weeks at #1 and counting.

And now, if anyone is wondering about good matchups according to SP+ next week:

  • #13 Kansas State @ #5 Texas
  • #17 Texas A&M @ #16 Ole Miss
  • #9 Notre Dame @ #21 Clemson
  • #20 Missouri @ #2 Georgia
  • #11 Oklahoma @ #33 Oklahoma State
  • #7 Washington @ #14 USC
  • #12 LSU @ #6 Alabama

Meanwhile, in the B1G…

Good shit, Iowa!

TRENDS (RAW SP+ SCORES)

Here are the weekly changes in the SP+ top 25 from the preseason through Week 9

SP+ OFFENSE AND DEFENSE (RAW SP+ SCORES)

Standard interpretation applies:

  • Top-right: Good at offense and defense
  • Bottom-right: Good defense, below-average offense
  • Top-left: Good offense, below-average defense
  • Bottom-left: Not so good at football

Recall that these are standardized coefficients, not raw SP+ scores. Each team’s offensive and defensive score represents the number of standard above/below the mean that category. For example, Michigan is 1.4 standard deviations above the mean from the average SP+ offense, and 2.1 standard deviations below the mean (this is good) from the average SP+ defense.   

Michigan, Georgia, and Texas are now the only two teams that are more than 1 standard deviation from the mean (in the right direction) on both offense and defense. Michigan and Iowa are basically bunk mates, and Iowa drew the bottom bunk. They better hope, like all of us, that Michigan doesn’t shit the bed this week.

Speaking of this weekend’s opponent, Purdue is…not Michigan.

SP+ RANKINGS BY CONFERENCE (RAW SP+ SCORES)

Horizontal lines are the average (mean) SP+ scores for each conference. Notre Dame is now excluded from the ACC. Thanks for the comments/suggestions on this. 

And because some readers asked for it, here’s the graph after conference realignment.

Anything you find interesting? Want me to explore in the future?

Thanks for reading. 

Go Blue!

Derek

November 1st, 2023 at 1:24 PM ^

Michigan, Georgia, and Texas are now the only two teams that are more than 1 standard deviation from the mean (in the right direction) on both offense and defense.

It's wild that OSU hits this on one of the dimensions, and it's not the offense. At least last year's defensive turnaround was buoyed by a great offense. Like the rest of you, I'm really looking forward to seeing it tested against a competent offense for the first time in the Big House.

BTW, USC is basically the Jon Kitna Lions, right? Lots of gun-slinging because the porous defense makes the whole game garbage time.

Bo Harbaugh

November 1st, 2023 at 1:32 PM ^

I am not a believer in OSU's defense. I don't believe they can stop the run consistently, and once they commit 6 to the box, they will get thrown on.

Against OSU, PSU, for some reason (Frames), decided to stop running despite getting 5-6 yard gains early in the game on 1st downs

McSomething

November 1st, 2023 at 2:17 PM ^

It was just more that those games lined up in the rock fight category as well. Indiana never looked like they were a threat to make it a real game, but Wiscosnin was able to hang around longer than they should've. And the Wisconsin offense making errors seemed to be why it wasn't closer as much as it was the OSU defense. 

Buy Bushwood

November 1st, 2023 at 2:13 PM ^

"Last year's defensive turnaround" was only extant against crappy competition.  Against the only two real opponents with diverse offenses, they gave up 45 & 42.  This makes perfect sense for an overrated risk-taker like Jim Knowles. He can swarm lesser teams who lack balance with his Big 12 schemes.  This is quickly punished against real offenses.  It's actually a pretty low-risk way to take on teams with huge talent deficits.  Because eventually you're going to get them in a loss and get them off the field.  But stopping the Michigans and Georgias requires you bleeding them down the field, trying to outlast them and not make mistakes.  This year is going to look a lot like last year, except that this year OSU doesn't have the offense to also bleed Michigan between the 20's.  

CMHCFB

November 1st, 2023 at 1:25 PM ^

Early candidate for post of the year, thx for putting in the work to detail all of that data.  
 

The post realignment P12 and B12 view is wild; Go Beavers. 

Logan88

November 1st, 2023 at 3:03 PM ^

There is actually nothing secretive about it. The posters handle includes "CMH" which is the callsign for the Columbus International Airport (I think it is now officially the "John Glenn Columbus International Airport" to be precise) and the local NBC affiliate in Columbus is WCMH. 

They did a pretty poor job of disguising their real affiliation. Of course, the obvious trolling in most of their posts was a pretty big giveaway as well.

Bo Harbaugh

November 1st, 2023 at 1:29 PM ^

S&P is great and I know it is one of the better tools for normalizing ranks based on output and opponent.  It also, of course, gets better as the season progresses as more data allows for a better model.

That said, it can be very reflexive, particularly in a situation like the B1G where there are some putrid offenses across the board, most specifically int the B1G West.  

I find it hard to believe that Iowa, UM, OSU and PSU are the 4 best defenses in the country.  At some point the eyes don't lie....the QB's in the B1G outside McCarthy are not good.

If UGA and Bama were in the B1G, I'm pretty sure they'd be grading out better than Iowa, OSU, PSU on defense and right there or better than UM.

mgoja

November 1st, 2023 at 1:47 PM ^

Is there any double counting going on or is this fully addressed through the normalization?  I.e., B1G defenses are generally much better than B1G offenses -- are the defenses ranked a bit higher than they should be because they are going up against so many bad offenses?  Or possibly/theoretically the opposite.

Either way it looks like Michigan has faced a few decent defenses so far and done very well, but hasn't yet had to defend against a living, breathing offense.

trueblueintexas

November 1st, 2023 at 1:48 PM ^

This is one of the reasons why I'm happy about the addition of the four west coast schools to the B1G. They will force the insulated midwestern schools to learn to play defense against good offenses and vice versa. This should make all teams in the B1G better. Steel sharpens steel. 

I don't think the SEC will get as significant benefit by adding Texas and Oklahoma. Those two schools already play a similar style of game as the rest of the SEC. 

LeCheezus

November 1st, 2023 at 4:41 PM ^

SP+ is opponent adjusted, but you can only "opponent adjust" so much.  If you fill a team's defensive data with nothing but putrid offenses, it can only filter that out to a certain point.  I do think Iowa's defense is top 10-15ish good, but this top 3 SP+ rating every year is absolutely a function of playing in the B1GW.  I can't say for sure, but I'd say the relative lack of NFL level talent probably means they are good but not one of the absolute best.  

I also don't think SP+ is good at accounting for an average of 20 minutes of garbage time per game, which is what Michigan has been averaging pretty much all year (outside of Rutgers).  The amount of data on Michigan's starting (or minimal rotation) defense and offense is much smaller than almost everyone else...which means we could rightfully be #1, but could also be getting that ranking on too little sample size.

leftrare

November 1st, 2023 at 1:53 PM ^

Thanks for doing this.  I'm hooked on it.  I refuse to pay Disney when the only thing I need from them that lives behind the paywall is SP+.  Connelly used to post so much more good stuff when he was independent and now it's all apparently gone.

 

mi93

November 1st, 2023 at 1:55 PM ^

I'm a little surprised osu is above ORE on the last chart.

And I also think the committee made osu #1 so that when Leon and friends finish "whoopin' dat ass", they can still back that ass into the playoffs as the 3 or 4 seed.  smh.

ShadowStorm33

November 1st, 2023 at 2:23 PM ^

Eh, it's just a reflection of OSU having played ND and PSU, which is probably the best set of two wins at the moment. Honestly, any non-champion is going to need a lot of help to get in. You can probably pencil in the B1G and SEC champions, so that's two out of four spots, and one-loss Big XII (Texas or OU), Pac-12 (UW or OR) and ACC (FSU) teams are getting in over a one-loss, non-champion OSU (or M)...

NittanyFan

November 1st, 2023 at 1:55 PM ^

Relevant to advanced analytics .....

These numbers are dated by a week (but wouldn't be dramatically different by adding a week either), and I originally saw this on reddit.  But I found it pretty hilarious:

(x-axis: a proxy for how good your offense is at its best.  y-axis: a proxy for how bad your offense is at its worst.  Higher x-values and lower y-values preferred)

M-Dog

November 1st, 2023 at 3:15 PM ^

This cries out for another version of it:  Hot = great offense.  Crazy = bad defense.

Because anybody that dates LSU (admittedly hot offense) is going to realize that LSU is batshit crazy (wacko defense) pretty quickly.

And USC?  We're talking restraining order.