Jump to content

S&P+ rankings: Overperformance or overcorrection?


satyanash

Recommended Posts

28 minutes ago, satyanash said:

So Oklahoma is favoured by ~15 points on a neutral site according to the S&P+ ratings from before this Friday. FEI also gives us just a 27% chance of winning in Arlington.

I don't think a one in four chance is unreasonable. It is a little low in my mind, but it isn't crazy. OU -15 is just silly.

Link to comment
Share on other sites

S&P+ predictions for all of the conference championship games:

Pac-12 Championship: No. 11 Washington 27, No. 17 Utah (+6) 25 (Nov. 30, 8:00 PM ET, FOX)

MAC Championship: Buffalo (-3.5) 26, Northern Illinois 21 (Nov. 30, 7:00 PM ET, ESPN2)

Big 12 Championship: No. 5 Oklahoma (-8) 43, No. 14 Texas 28 (Dec. 1, 12:00 PM ET, ABC)

Sun Belt Championship: Appalachian State (-16.5) 39, UL-Lafayette 21 (Dec. 1, 12:00 PM ET, ESPN)

SEC Championship: No. 1 Alabama 32, No. 4 Georgia (+13.5) 29 (Dec. 1, 4:00 PM ET, CBS)

AAC Championship: No. 8 UCF (-3) 38, Memphis 28 (Dec. 1, 3:30 PM ET, ABC)

Conference USA Championship: MTSU (-1.5) 26, UAB 24 (Dec. 1, 1:30 PM ET, CBSSN)

Big Ten Championship: No. 6 Ohio State (-14.5) 38, No. 21 Northwestern 18 (Dec. 1, 8:00 PM ET, Fox)

ACC Championship: No. 2 Clemson 41, Pittsburgh (+28) 16 (Dec. 1, 8:00 PM ET, ABC)

MWC Championship: No. 25 Fresno State (+3) 31, No. 22 Boise State 26 (Dec. 1, 7:45 PM ET, ESPN)

Link to comment
Share on other sites

11 minutes ago, satyanash said:

S&P+ predictions for all of the conference championship games:

Pac-12 Championship: No. 11 Washington 27, No. 17 Utah (+6) 25 (Nov. 30, 8:00 PM ET, FOX)

MAC Championship: Buffalo (-3.5) 26, Northern Illinois 21 (Nov. 30, 7:00 PM ET, ESPN2)

Big 12 Championship: No. 5 Oklahoma (-8) 43, No. 14 Texas 28 (Dec. 1, 12:00 PM ET, ABC)

Sun Belt Championship: Appalachian State (-16.5) 39, UL-Lafayette 21 (Dec. 1, 12:00 PM ET, ESPN)

SEC Championship: No. 1 Alabama 32, No. 4 Georgia (+13.5) 29 (Dec. 1, 4:00 PM ET, CBS)

AAC Championship: No. 8 UCF (-3) 38, Memphis 28 (Dec. 1, 3:30 PM ET, ABC)

Conference USA Championship: MTSU (-1.5) 26, UAB 24 (Dec. 1, 1:30 PM ET, CBSSN)

Big Ten Championship: No. 6 Ohio State (-14.5) 38, No. 21 Northwestern 18 (Dec. 1, 8:00 PM ET, Fox)

ACC Championship: No. 2 Clemson 41, Pittsburgh (+28) 16 (Dec. 1, 8:00 PM ET, ABC)

MWC Championship: No. 25 Fresno State (+3) 31, No. 22 Boise State 26 (Dec. 1, 7:45 PM ET, ESPN)

no way is OU holding us to 28. we only played for 3 quarters the first game and scored 48.

Link to comment
Share on other sites

  • 2 weeks later...

I‘m curious how S&P+ does during bowl season vs. regular season. I would think not nearly as well because of all the factors in bowl season that it just isn’t set up to account for (coaches/assistants leaving or on their way out the door, players sitting out, teams disappointed in their finish/bowl who just kinda check out, etc.).

Link to comment
Share on other sites

4 hours ago, satyanash said:

S&P+ currently has the Sugar Bowl as Georgia 40, Texas 22.

40 seems really high.  Georgia's offense scored 29 on MU and 27 against Auburn.  10 of the 36 points vs Florida were off turnovers.  UGa's offense struggles to break 30 against any defense with a pulse.  They will likely need big plays by the special teams or defense to reach 30, much less 40.

Link to comment
Share on other sites

I can't see the detailed advanced stat profiles, but I agree that 40 points seems too high, especially since only 2 teams eclipsed that mark all season against us (OU and WV) and it doesn't seem that Georgia's offense, despite being 3rd in S&P,  is explosive enough or more importantly plays at the pace necessary to challenge that mark the way that OU and WV could. If we lose by 18, it would be in more of the 28-10 variety, not 40-22.  If someone gave me TX +18 in a bowl pool, I'd take that all day as one consistent trend is that the S&P+ model has generally underrated TX ( and Herman teams in general for that matter) as an underdog the last couple of years. Even Connelly only half believes his model - 

  • Sugar Bowl: No. 5 Georgia (-12) 40, Texas 22 (Jan. 1, 8:45 p.m. ET, ESPN)

Big 12 Championship aside, Tom Herman teams are generally very good as an underdog; that’s really the only reason to think this one ends up close. (Man, I don’t like these NY6 matchups very much.)

Link to comment
Share on other sites

  • 3 weeks later...

Fuck those stupid ass rankings. I'm sorry but I refuse to believe such powerhouses as Fresno state, App state and Memphis rank ahead of us in anything. For a poll that touts being predictive of the future, why doesn't it take into account new coaches? Handful of teams face that reality for next year. Anyone wanna wager on da U and WVU having worse records next season?

Preseason polls are dumb but this takes it to a new level. 

  • Like 1
Link to comment
Share on other sites

24 minutes ago, HtownHorn said:

It's what happens when you win 10 games, and not have an offensive play of 50+ yards all season.

And have down-to-the-wire games with inferior opponents week after week. 

BC himself points out time and time again that Herman is a magician at preparing and delivering in the spotlight games, when his team is the underdog, but week-to-week and when they are the favorite they consistently underperformed. 

This shouldn't be anything new to anyone who has watched Texas play the last two years, or followed Herman's time at Houston, that is Herman's (too this time, very short sample size as a HC) team MO. 

Trying to statistically model a team that is so schizophrenic is not easy.  

Edited by TheYoungHorn
Link to comment
Share on other sites

11 minutes ago, TheYoungHorn said:

And have down-to-the-wire games with inferior opponents week after week. 

BC himself points out time and time again that Herman is a magician at preparing and delivering in the spotlight games, when his team is the underdog, but week-to-week and when they are the favorite they consistently underperformed. 

This shouldn't be anything new to anyone who has watched Texas play the last two years, or followed Herman's time at Houston, that is Herman's (too this time, very short sample size as a HC) team MO. 

Trying to statistically model a team that is so schizophrenic is not easy.  

Some of the schizophrenic was due to injuries costing Texas key players. If Sam plays all game against BU, Texas likely blows them out. If Brandon Jones doesn't injure his ankle fucking around before the West Virginia game, Texas likely wins that game as well. If Boyd and Davis don't go full retard before the OSU game, Texas likely wins that game as well. Texas's top 30 players can compete with all but 1-2 teams in the country. After that, it gets dicey, but Herman is stacking talented classes together to build quality depth.

Link to comment
Share on other sites

Fuck those stupid ass rankings. I'm sorry but I refuse to believe such powerhouses as Fresno state, App state and Memphis rank ahead of us in anything. For a poll that touts being predictive of the future, why doesn't it take into account new coaches? Handful of teams face that reality for next year. Anyone wanna wager on da U and WVU having worse records next season?
Preseason polls are dumb but this takes it to a new level. 


That’s a postseason ranking, not a preseason one.
Link to comment
Share on other sites

5 hours ago, Machinator said:

We don't make sense to Bill's system, and that confounds him.

We really are an outlier to him. I think his system works really, really well for pretty much everyone else, which is making me believe Herman is some kind of young warlock, an orange wizard in training, if you will. He consistently slays dragons with dark magic, does just enough to beat most bad to mediocre teams, but is occasionally susceptible to some kind of kryptonite against others in that category. 

I think that one loss to Maryland dragged us down all year long like a block of cement tied around our neck. 

Actually, Okie State confounded him all year long too....Considering he is a Big 12 grad, you'd think he'd have a better handle on this conference by now, but he doesn't. We are the weird P5 conference. Nothing makes sense here. Who is your commanding officer?

 

Edited by MaybeACoordinator
Link to comment
Share on other sites

Comparison of Tom's actual number of wins versus S&P+'s 2nd Order Wins for each of his seasons as a HC:

2015: 13 (10.2)

2016: 9 (8.8)

2017: 7 (7.1)

2018: 10 (8.2)

Not exactly a consistent pattern of "gaming" S&P+.  Keep in mind that every season there are a number of teams with a substantial (e.g., 1.5 or greater) difference between the number of actual wins and 2nd order wins.  

Link to comment
Share on other sites

Success rate is what drives S&P+ though, not the score.  If you have a high marginal success rate against your opponent, but lose due to turnovers or not maxing out your points/opportunity, S&P+ will still grade you out well for that game.  I think it would be fair to describe Herman's offensive system as "plodding" at times, particularly when we have a lead, but I believe we are still trying to be successful on each down when we have the ball.  

Link to comment
Share on other sites

3 minutes ago, Fozzz said:

Success rate is what drives S&P+ though, not the score.  If you have a high marginal success rate against your opponent, but lose due to turnovers or not maxing out your points/opportunity, S&P+ will still grade you out well for that game.  I think it would be fair to describe Herman's offensive system as "plodding" at times, particularly when we have a lead, but I believe we are still trying to be successful on each down when we have the ball.  

Explosiveness is a big component. I think Texas this year and his system show that maybe it's too big of a component. Perhaps drive success rate should be a bigger factor and explosiveness a smaller factor in the overall evaluation.

And once again nobody has been able to supply an explanation for how his system arrives at the overall unit number. Texas offense this year:

image.png

We are #44 in drive efficiency, #111 in explosiveness, #98 in rushing, #43 in passing, #69 on standard downs, and #75 on passing downs. Somehow that adds up to #26 overall. That makes no sense whatsoever. There are some aspects that aren't in that chart but we're actually bad at those things this year (red zone, etc.).

Link to comment
Share on other sites

I think in the newest formula he uses efficiency matters much more than explosiveness, which has been reduced in importance since his discovery that IsoPPP is mainly noise.  I do agree that I wish Bill would just provide the formulas he's using so we would know what weights he's using.    

As for our offensive rating, the fact that we have an overall rating of 26th with an overall success rate of 31st and an IsoPPP of 115th shows how little a role explosiveness now plays in determining offensive S&P+.  This season Texas' offense did a good job of staying on schedule and took care of the ball (all in keeping with Herman's offensive strategy), but was terrible at generating big plays.  Funnily enough, Herman's offensive strategy appears to highlight the skills that S&P+ favors.

Edited by Fozzz
Link to comment
Share on other sites

Our success rate is 44th. There's no way that those components add up to a #26 overall ranking, something is off and this same thing happened previously in 2016 under Strong.

Big image spoilered because I realized I didn't want to put up a link to the other site:

image.png

Edited by Huckleberry
Link to comment
Share on other sites

That's weird, as the components are different than what's shown in the Google doc that Bill refers to on twitter when he updates the stat profiles.  I wonder if the components have been updated on Football Outsiders.  Here's what Bill has in his Google Doc:

image.png

 

It would appear he's putting a lot of weight on the marginal efficiency components, which are new for this year.   

Link to comment
Share on other sites

Being good/decent at several things but not great at anything would generate a much higher ranking if his system were geometric as opposed to linear.

I don't know if that makes sense for other teams though.

Edited by JBJ
Link to comment
Share on other sites

  • 3 weeks later...

Still 12 spots behind aggy. They also played a lot of close games against mediocre to bad P5 teams. The big difference seems to be they played 3 G5 or lower teams and won them handily. We played 1 G5 and didn't look great. I am going to guess that Connelly's system claims to be adjusted, but in the end a conference that averages 3 G5 or lower games per team (and of course, on average, destroys those teams) is going to get a boost once all the cross-pollination occurs. I also think the high rankings for UCF, Fresno St, App State, and Utah State suggests beating up on weak teams is the best way to improve S&P.

Link to comment
Share on other sites

He should be grateful for Texas and Herman. Knowing that one's model needs tweaking is typically no where near as useful as getting some specific examples of related prediction deficiencies. Close games are inherently harder to predict, but teams that tend to get into (or tend to win or tend to lose) close games may be manifesting some previously occult process. Sort of like the "don't outrun the bear, just outrun your slowest fellow camper" effect, perhaps. We had dominating performances twice against teams that were predicted to beat us by double digits. Both times we let them back close on the scoreboard but won. Those games may offer an opportunity for modelers to learn from Texas and vice versa.

Link to comment
Share on other sites

Overall, the biggest issue with S&P+ as it applies to UT is that it seems our coaches actively tried to bleed clocks in exchange for our team playing worse. Anytime we got a two score lead or more in the second half, the offense seemed to go into a shell where it would go RB run 1st and 2nd, then incomplete pass. Our OL was not a good run blocking unit, so when a defense could expect run, they usually shut us down, until roughly the last drive of the game, we'd get creative again and also run Sam and viola, we'd pull out a 6 minute drive and ice the game. On defense, Orlando would quit blitzing, start dropping a bunch of guys and offenses would start to carve us up. Look at the UGA game, when we went into prevent-type defenses, they tore us up, but the one late drive we stayed aggressive, we forced a 3 and out. 

Anyways, I think it's a flaw by the coaches and I hope they quit doing it, but I also think S&P+ overpenalized us for it because it doesn't take into account the idea that coaches might actively sacrifice team performance for bleeding the clock. Conversely, when we needed to show up/didn't have a huge lead, we outplayed almost every team, including OU in the Cotton Bowl and Georgia. I realize it doesn't work this way, but if you could filter out stretches of games where our coaches got uber-conservative on both sides of the ball with a big lead, our S&P+ would likely be more accurate. Also, this is different then just saying, eliminate all our bad stretches of play, because you'd still leave games like Maryland and OSU in the data.

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...