ESPN College Football - Computer Ranking - Top 10
The attached article caught my eye. The whole idea of ranking - and subjectivity vs objectivity to assess teams while selecting the teams for the CFP - will be an ongoing conversation.
The link follows (it is not paywalled) and, was published AFTER Saturday's games.
Here's the latest top 10 from ESPN's computer model:
- Georgia
- Alabama
- Ohio State
- Michigan
- Tennessee
- Texas
- Penn State
- Utah
- Clemson
- TCU
Yes, after Michigan's victory in Columbus, the eSECpn computer algorithm still has the Buckeyes ranked in front of the Wolverines. And, there are only three SEC teams in the Top Five.
As an engineer whose career has been in technology, I'd be very interested in seeing the algorithm that determined these rankings.
My immediate reaction is essentially - incredulous.
November 29th, 2022 at 10:52 AM ^
When it comes to data and algorithms like this - garbage in, garbage out. Clearly they’re not using the right datapoints as this is incredulous indeed
November 29th, 2022 at 11:07 AM ^
Algorithm clearly didn't expect (and doesn't going forward) that Michigan would be able to win a game against OSU by breaking the vast majority of its tendencies (explosives). Down by down, without Corum, we struggled on offense until Q3. The model just isn't built to handle UM's approach to football.
November 29th, 2022 at 11:15 AM ^
The model just isn't built to handle UM's approach to football.
Neither was Ohio State....amirite????
November 29th, 2022 at 11:46 AM ^
To be fair, no one is.
November 29th, 2022 at 11:29 AM ^
Punishment for the new TV deal. I wonder how they will handle game day in 2 yrs.
November 29th, 2022 at 11:46 AM ^
Ah hell, and the very first comment says what I posted 39 comments down. I'm ashamed.
November 29th, 2022 at 11:58 AM ^
Algo:
SEC = B1G + x$(Saban slurp ratio)
OSU = UM * (10 year Points history OSU/UM)
November 29th, 2022 at 2:00 PM ^
I fully believe ESPN's power rankings are more or less bullshit, but it looks like the MSN article is using an out-of-date ranking that does not account for Saturday's games. If you go to the ESPN Power Rank page, it shows this list hasn't been updated since Nov 21, 2022 (see bottom left-hand corner): https://www.espn.com/college-football/fpi
November 29th, 2022 at 10:53 AM ^
Makes sense. Replace Ohio State with Florida and Michigan with LSU and I think it would be perfect.
November 29th, 2022 at 10:53 AM ^
LOL. Imagine the conversation in trying to convince a team that this ranking is spot on; how on earth did this get published. ESPN is clearly shit.
November 29th, 2022 at 10:54 AM ^
I don't know shit about fuck, but it's pretty clear that these rankings are a mess.
November 29th, 2022 at 11:30 AM ^
i know a lot about fuck, but not shit about algorithms.
still pretty sure this fucking algorithm is broken.
November 29th, 2022 at 12:40 PM ^
They know it's broken. LSU should have slid into the #6 spot...they're still working on the technical issues.
November 29th, 2022 at 10:55 AM ^
A perfect encapsulation of why metrics are often bullshit and why we should never let them take over sports rankings. We've seen the baseball sabermetrics for years tout certain players as having more value than others only to see those players never win a championship or really be a factor at all.
Whatever this algorithm is, it's garbage.
November 29th, 2022 at 10:55 AM ^
Having SEC in your letterhead apparently earns a hefty weight in their model.
November 29th, 2022 at 10:57 AM ^
All I had to see is Texas at 6 to almost laugh myself to death.
November 29th, 2022 at 11:09 AM ^
THAT'S all you had to see? There are four other "thats" above Texas in this nonsense.
Totally agree that anyone who looked at this and said, "Yeah, let's put this out there," while still thinking they would have an ounce of credibility after is a...well, I guess that would be par for the course, wouldn't it?
November 29th, 2022 at 11:12 AM ^
But Saban says their losses were close.
November 29th, 2022 at 11:22 AM ^
And, in a news conference - Saban noted that Bama's two losses were to teams ranked in the Top 10.
Of course, those were "Top 10" SEC teams when they played them. Those two teams records are now: Tennessee at 10-2, and LSU at 9-3. Tennessee is ranked 8th in both the AP and Coaches polls while LSU is ranked 11th (AP) and 13th (Coaches).
November 29th, 2022 at 11:18 AM ^
It's all ridiculous but Texas at 6, with an 8-4 record and at best a borderline Top 25 team in the real world, is so egregious it really makes the entire thing hilarious.
November 29th, 2022 at 11:07 AM ^
Just goes to show the lengths that they’d go to get Bama and OSU in the playoffs and could still happen. If TCU and Usc lose I could easily see GA vs OSU AT 1 and 4 and us vs Bama at 2 and 3. I’m sure the assumption would be that all is right in CFP with a Bama Ga natty.
November 29th, 2022 at 11:07 AM ^
If your algorithm has OSU over Michigan after Michigan just beat them by 3 possessions in Columbus, your algorithm sucks. Don't come back and say iT's a mOdEl. If it can't even discern that Michigan is better than OSU then it's not worth considering
November 29th, 2022 at 12:02 PM ^
just hacked ESPN computer and found out the algorithm has a name. It's called
"The ESPN Super Accurate How to Get Bama in the Playoffs so CFP Committee Please Look at This Model and Use it as a Guideline Moving Forward Pretty Please with Sugar on Top" algorithm
November 29th, 2022 at 11:08 AM ^
Wow what a bunch of clowns. Probably should have kept this list to themselves.
November 29th, 2022 at 11:10 AM ^
Points for each game you play against an SEC opponent + points for each ESPN watcher in your fanbase + points for each ABC/ESPN announcer who is an Alum, oh and I guess we'll throw a point in for each victory.
November 29th, 2022 at 11:11 AM ^
This is straight FUD from ESPN. Expect ESPN to start pushing some unhinged-from-reality “two SEC teams in playoffs” narrative, and they will start pointing back to this batshit computer ranking. And way too many people will eat this shit up.
November 29th, 2022 at 11:11 AM ^
If you don't know what variables are used to obtain such results, it's not really worth a conversation. Bill Connelly says, 'I take this this and this into account' and we stare at it, examine it over time, he tweaks it, and the measure comes to be accepted as fairly accurate or not. If this were about latent potential as possessed by these teams in re: their recruiting, etc., maybe. . . What Harbaugh and co. have made clear is that they can put together teams that are more than the sum of their parts. That's why you play in the first place.
November 29th, 2022 at 11:11 AM ^
and people wonder why ESPN is going in the tank.
November 29th, 2022 at 11:50 AM ^
Do they, though?
November 29th, 2022 at 11:14 AM ^
It’s not a ranking of what you’ve done though.
This happens every time FPI or SP+ is brought up for discussion.
Resume-based rankings (AP Poll, CFP Poll, etc.) are fundamentally different than predictive models (FPI, SP+).
This model likely accounts for the fact that nearly all of Michigan’s offense came in about 5 plays and would not expect it to happen again (but that doesn’t mean it couldn’t or wouldn’t).
That said, I would be curious what goes into calculating FPI (SP+ is pretty transparent), but ESPN does keep the lid on it.
November 29th, 2022 at 12:00 PM ^
M's explosive plays were a function of M's previous success with running and mounting sustained drives and OSU trying to stop that. Seems like a "predictive" model that effectively discounts those explosive plays as flukes is off.
November 29th, 2022 at 12:03 PM ^
And the reason why Michigan's offense came in about 5 plays is because OSU was loading the box to stop the run. If OSU plays it normal, then Michigan grinds out 5-6 yards per rush. I don't think the outcome would be much different, but the game would've played out much differently.
November 29th, 2022 at 11:17 AM ^
I think the flaw in the model is apparent from the results - record isn't taken into account. That is, there is no penalty for losing a game. Most important thing seems to be strength of schedule (1. Bama 2. Tenn 3. Texas 4. OSU 5. UGA).
Where Baylor is ranked would be informative, because they are 6-6 against a top 15 schedule.
November 29th, 2022 at 11:20 AM ^
Some bad =if statements in that math.
November 29th, 2022 at 11:27 AM ^
Forget the algorithms, you just know the Committee is jonesing to get Bama that 4-seed next week after Kansas State wins Saturday.
November 29th, 2022 at 11:27 AM ^
Well the algorithm is written in crimson crayon by a guy named Feet Pinebottom so digging deep may take you past your answer.
November 29th, 2022 at 11:27 AM ^
Putting Alabama #2 tells you all you need to know. I would like to see the week to week of this computer ranking. If this is the first week of the ranking, then obviously, the data was skewed to get the result they want. It's ESPN. We know they have no objectivity.
November 29th, 2022 at 11:29 AM ^
This list is, of course, hot garbage. And it shows that certain people (including some prominent analysts) who have said, "The CFP Committee stinks! You should use an unbiased computer!" are just a dumb -- computers are built by people. They can give you just as bad "unbiased" results (e.g., I am sure the computers absolutely love OSU because of margin of victory).
As someone said, garbage in, garbage out. Stick with the committee with a clear setup (e.g., in new system, all Power 5 conference champs make the playoffs) -- at least then you have someone you can hold accountable.
November 29th, 2022 at 11:30 AM ^
I have absolutely no idea how any objective tool could look at Alabama and Texas this season and considering them a top-10 teams. I'm sure their model does that thing all bad models do where it relies on a series of assumptions and they become self-referencing and fulfilling; Alabama beat a top-6 Texas team close so that means Alabama is good and Texas lost a close game to a top-6 Alabama team so that means Texas is good and because of that Alabama's close win over Texas makes them a top 6 team so that's a good loss for Texas to a top-6 Alabama team so Texas deserves credit...and all the way down. Do the same with Tennessee and Alabama and you see it again.
It's funny to me that ESPN brought on Bill Connelly and his far superior SP+ system and yet still trot out this shitty copy-cat system they created on a lunch break for some inexplicable reason.
November 29th, 2022 at 11:30 AM ^
Let's check in with the ESPN data analysis guru...
November 29th, 2022 at 11:47 AM ^
I hate to say it, but my simian brother is correct.
November 29th, 2022 at 12:53 PM ^
ESPN meeting this morning, with dude explaining the algorithm
November 29th, 2022 at 12:59 PM ^
:-) we aren't trainable.
November 29th, 2022 at 11:34 AM ^
Sagarin computer rankings have Ohia Michigan 2-3. Not just ESPN. Shows the flaws in these models when coming off an 22 point Michigan win. Also curious about SOS. Ohia played ND. But Michigan played Illinois.
November 29th, 2022 at 11:36 AM ^
At some point, people need to actually watch the games. Feng downplays Michigan's win because we won with some 'explosive plays', but doesn't acknowledge that 1 of OSU's 2 touchdowns came on an explosive play and that Michigan's defense held OSU's "elite offense" to only one TD in however many red zone trips they had. Also, our explosive pass plays only happened because OSU sold out to stop the run. If they try to prevent the pass, then we'd just run right over them like we did at the start of the 2nd half.