Jump to content

S&P+ rankings: Overperformance or overcorrection?


satyanash

Recommended Posts

  • 2 weeks later...
On 7/8/2019 at 6:46 PM, GabrielsHorn said:

FPI updated their rankings:

http://www.espn.com/college-football/statistics/teamratings

Any system that has UCLA and Tennessee ahead of us should be thrown out. 

Something is definitely wrong with these rankings. There are a few SEC teams behind us. Gotta get that fixed Bill.

Edited by B00M
Fuck the SEC
  • Like 1
Link to comment
Share on other sites

On 7/8/2019 at 6:46 PM, GabrielsHorn said:

FPI updated their rankings:

http://www.espn.com/college-football/statistics/teamratings

Any system that has UCLA and Tennessee ahead of us should be thrown out. 

Wow. I've actually liked FPI as far as statistical rankings go, but there are some real head scratchers there. Tennessee and UCLA as you mentioned. Mizzou at 19. Auburn at 9. Some serious SEC inflation there. I think Oregon is low at 10 as well. And even removing the burnt orange shades, Texas at 24, below Iowa and the rest of the teams listed above, is lulzy. Oregon plays Auburn in Jerryworld week 1. I am thinking Oregon will win and manage to still be ranked below Auburn, even though they're 9-10 now.

Link to comment
Share on other sites

13 minutes ago, gmr548 said:

Wow. I've actually liked FPI as far as statistical rankings go, but there are some real head scratchers there. Tennessee and UCLA as you mentioned. Mizzou at 19. Auburn at 9. Some serious SEC inflation there. I think Oregon is low at 10 as well. And even removing the burnt orange shades, Texas at 24, below Iowa and the rest of the teams listed above, is lulzy. Oregon plays Auburn in Jerryworld week 1. I am thinking Oregon will win and manage to still be ranked below Auburn, even though they're 9-10 now.

Honestly, any metric that values returning production is going to rate us low because we're turning over the whole defense except the safeties.

Link to comment
Share on other sites

4 minutes ago, satyanash said:

Interesting that ESPN's FPI (7.7 projected wins) is significantly higher on us than S&P+ (6.7 projected wins). The gap is even higher for Iowa State.

I know S&P+ loves returning production and margin of victory, both of which we don't look great on. I wonder whether FPI has those calibrated lower.

Link to comment
Share on other sites

42 minutes ago, texifornia said:

 

Notwithstanding all the critical flaws with Connelly’s system, If Sam doesn’t get banged up those last few weeks and the defense stayed healthier weeks 11-15, that chart would  probably show a pretty nice growth curve.

also, ISU’s chart is pretty cool. Really puts into perspective what an incredible job Matt Campbell has done.

D_waGtzXsAIZBOU?format=jpg&name=medium

Link to comment
Share on other sites

  • 4 weeks later...

Geez, man, being required to meet output quotas is killing his ability to think things through.

Does he really think average all time rating without accounting for years in the ratings is the best way to list programs? He wouldn't need an exclamation point after a team in his ratings if he'd accounted for stuff like that.

Link to comment
Share on other sites

5 hours ago, Huckleberry said:

Geez, man, being required to meet output quotas is killing his ability to think things through.

Does he really think average all time rating without accounting for years in the ratings is the best way to list programs? He wouldn't need an exclamation point after a team in his ratings if he'd accounted for stuff like that.

In general terms some of the team charts are fun.

Nebraska's questionable blueblood-ness being based on like two people: 

And his own Mizzou Tigers being schizo af

 

Link to comment
Share on other sites

Looking at the SP+ preseason page at ESPN (I feel a little dirty for giving them the click), BC appears to include some marginal input in his formula:

W/L from 5 years ago

Freshman class recruit rankings

While excluding a few data points that might be relevant:

HFC/coordinator turnover

3rd year class recruiting ranking/retention

Key offseason personnel losses to injury

As a result, he has Texas, WSU, and Nebraska at 31, 34, and 39, while clustering Auburn, Mississippi St, and aggy in the 8-11 range. I am not a gambling man in general, but would happily bet that he has this wrong. I’ll pay 5 to 1 that at least one of those lower-ranked teams finishes above at least one of his list of higher-ranked ones, despite the gulf of 20 spots between the clusters.

Link to comment
Share on other sites

6 minutes ago, Huckleberry said:

Usually they're in Missouri gear, but every once in a while one of the Connellys will wear another school's gear. Always interesting to see which school it is.

image.png.d4640474921a8ad3c421e11e39932dc1.png

He's from Oklahoma and went to Mizzou. He's quite open about that.

He's less open about what that means for his feelings about Texas.

Link to comment
Share on other sites

12 minutes ago, Magus Ossis said:

Looking at the SP+ preseason page at ESPN (I feel a little dirty for giving them the click), BC appears to include some marginal input in his formula:

W/L from 5 years ago

Freshman class recruit rankings

While excluding a few data points that might be relevant:

HFC/coordinator turnover

3rd year class recruiting ranking/retention

Key offseason personnel losses to injury

As a result, he has Texas, WSU, and Nebraska at 31, 34, and 39, while clustering Auburn, Mississippi St, and aggy in the 8-11 range. I am not a gambling man in general, but would happily bet that he has this wrong. I’ll pay 5 to 1 that at least one of those lower-ranked teams finishes above at least one of his list of higher-ranked ones, despite the gulf of 20 spots between the clusters.

That's going to be offset by the fact all three of that high cluster play each other - if the worst of the three has a bad time it'll naturally get knocked down.

Link to comment
Share on other sites

13 minutes ago, texifornia said:

He's from Oklahoma and went to Mizzou. He's quite open about that.

He's less open about what that means for his feelings about Texas.

lulz, dude's a major homer and everyone knows it. For the most part the homerism had not infected his actual analysis until recently. Previously it had been limited to the text of his columns that accompanied the analysis.

But how he's done stuff like adding which conference a team is in to his numbers. You and I both know that he would have never done that before Missouri was in the SEC. Now it seems like a great idea to him. Hell, he owns and wears a "Missouri" shirt that's really just a huge SEC logo on a yellow shirt.

  • Like 1
  • Fuck You 1
Link to comment
Share on other sites

25 minutes ago, texifornia said:

That's going to be offset by the fact all three of that high cluster play each other - if the worst of the three has a bad time it'll naturally get knocked down.

No it won't, he's tweaked his system so that they will stay high. Besides, three highly ranked teams all playing each other get knocked down in human polls, not in computer systems.

Edited by Huckleberry
  • Like 1
  • Fuck You 1
Link to comment
Share on other sites

One other thing I failed to mention on the other thread is this attitude that some of the advanced ratings guys have been taking acting like Texas is overrated by 20+ spots in the human polls compared to all the advanced systems.

Sagarin - #13
Massey - #16

It's not like every computer system hates Texas.

  • Like 1
Link to comment
Share on other sites

https://www.espn.com/college-football/story/_/id/27461638/college-football-conference-previews-storylines-using-sp+

 

Yes, Texas is 31st. FPI has the Longhorns just 26th, too. Tom Herman's propensity for overachieving in big games but underachieving in the smaller ones leaves his teams ripe for upsets and forever drags them down from a numbers perspective. Texas fans have spent the offseason yelling things like "Yeah, we only beat Tulsa by a touchdown, but we physically dominated them! We were never threatened!" That's great. Clemson and Alabama would have beaten Tulsa by 48 without turning to the second page of the playbook. To look like title contenders, Texas has to treat every game like a Big Game, not just two or three.

Link to comment
Share on other sites

2 hours ago, bigcigar said:

https://www.espn.com/college-football/story/_/id/27461638/college-football-conference-previews-storylines-using-sp+

 

Yes, Texas is 31st. FPI has the Longhorns just 26th, too. Tom Herman's propensity for overachieving in big games but underachieving in the smaller ones leaves his teams ripe for upsets and forever drags them down from a numbers perspective. Texas fans have spent the offseason yelling things like "Yeah, we only beat Tulsa by a touchdown, but we physically dominated them! We were never threatened!" That's great. Clemson and Alabama would have beaten Tulsa by 48 without turning to the second page of the playbook. To look like title contenders, Texas has to treat every game like a Big Game, not just two or three.

One can argue about the point of models, but their general purpose is predictive. Again, are they more to approximate scores or identify which team will win? Many here likely feel that winning by 1, while less satisfying than by 21, is still winning. They likely take exception to the philosophy strongly implied by Herman's statement that "you only need to win by one," namely that shooting for marginal success is not an unnecessarily risky proposition. That said, the difference between winning by 1 and losing by 1 is much bigger than the difference between winning by 1 and winning by 11. People go crazy when a last-second FG lifts them from, say, down 24-25 to up 27-25. A lot of models don't much care. Going 6-6 with 6 comfortable wins and 6 razor-thin losses is not remotely as good an outcome as squeaking out an undefeated season. If those models seek to predict, e.g., whether Texas will play in (and perhaps win) important post-season games, then they were objectively failures in that regard last year. They appear on track to fail similarly this year.

I don't think anyone here is arguing that Texas was as good as Clemson or Alabama last year (or now for that matter). Nor are people arguing that the models bear no relation to the system they seek to model and predict. Rather, especially with respect to SP+, the case has been made that (1) sound empirical scientific and statistical technique has not been used to develop the models, (2) this (or something) has resulted in some peculiar failings of the model-- of interest here because Texas is one of the teams whose winning it seems to predict least accurately, and (3) the author and advocate of the model has behaved in a manner to suggest that he both senses to a degree the need to improve his creation as well as that he refuses to concede publicly any of its deficiencies. To the contrary, he makes statements to suggest that the reality he tried to model was somehow failing to conform to his model rather than the converse. I think virtually all here would agree that to look like a title contender, Texas needs to stop losing to the Marylands of the world and to stop letting the Tulsas get back in the game; to suggest otherwise is to miss the point of the criticism leveled at BC and his SP+. Nota bene, however: more important than looking like a title contender is becoming one-- if Texas wins every game one the schedule by 1 point, they will be one.

Who will bet me straight up on his predicted W/L total for Texas if I get the over? Any takers?

  • Like 2
Link to comment
Share on other sites

58 minutes ago, Magus Ossis said:

Who will bet me straight up on his predicted W/L total for Texas if I get the over? Any takers?

That’s a lot of words.

I wouldn’t touch the 6.9 wins and don’t think anyone but aggy thinks it’s a remotely accurate prediction. That’s absurd and shows a significant flaw in his model (as you explained).

Bill is getting quite defensive about the issue. It’s going to be fun reminding him of his posturing throughout the season.

  • Like 1
Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.



×
×
  • Create New...