Jump to content

S&P+ rankings: Overperformance or overcorrection?


satyanash

Recommended Posts

1 hour ago, Hank Chinaski said:

There are 2 big differences between this and baseball stats that immediately jump to mind:

(1) baseball provides very large sample sizes. This S&P+ is going to lack power because it is based on a bunch of teams that have played 6 games. 6. Things like run differential in baseball are pretty good predictors of how good a team actually is, but not after 6 or 12 or 18 games. 

(2) strategy in baseball doesn't change with game situations as dramatically as it does in football. In general, the hitter is always trying to hit the ball hard and the pitcher is always trying to get him out. Of course there are times when making contact or pitching around a hitter are prioritized, but that, again, is where the sample size comes in. In football, sometimes the goal is to run clock, sometimes the goal on defense is just to prevent quick scoring, etc. 

I'm not saying S&P+ doesn't tell you anything, but it seems more like it is more useful as a betting tool than a "ranking" system in terms of ranking how good the teams actually are. 

Thank you, this pretty much sums it up. 

Link to comment
Share on other sites

5 minutes ago, Pato del Muerto said:

And that’s why it thinks we only have about a 30% chance to win in Stillwater. Anyone feel that prediction is fair?

No. It's dumb as hell. We might lose but there's no way we only have 30% chance of beating Oklahoma State in Stillwater assuming Ehlinger is healthy. Oklahoma State has been dominated 3 out of the last 4 weekends by Iowa State, Texas Tech, and K-State. 2 of those were at home.

30% chance to win in Stillwater is laughingly bad and an indictment of how dumb S&P is. 

 

Edited by texasstrong12
  • Like 1
Link to comment
Share on other sites

1 hour ago, NeverMarryAStripper said:

 

A predictive poll that shows that we would lose to 1/3 of the teams we beat seems like it sucks

I think this is math vs human perception. We know all to well the perception of a team that wins by couple of points by scoring in the last seconds. The whole post game is about how this team is a winner, their players are amazing, their future is bright, etc. and the losers are the opposite; whereas in reality, it is a statistical tie. This dichotomy is so pronounced that I find it very amusing watching games when I don't have any rooting interest.

But for the computer, not only is it a statistical tie, but the loser may have, in quantifiable terms, played better than the winner. But in the human poll, they lose 10 spots and the winner gains 5 spots, and the division is magnified beyond reality.

All S&P is saying is that given a Baylor or Tulsa or KSU or God forbid, Maryland, we are a one score toss-up, or in other words, a statistical tie. Also, given an ISU, OSU or TTech, we are predicted to be, surprise-surprise, a toss-up. The difference of some toss-ups being above us is secondary effects.

  • Like 2
Link to comment
Share on other sites

1 minute ago, AeroHorn said:

But for the computer, not only is it a statistical tie, but the loser may have, in quantifiable terms, played better than the winner.

Here's your problem. There is only one quantifiable term in which a team can play better than another team:  scoreboard.

If you win, you are a winner. If you lose, you are a loser. Any other metric is bullshit.

  • Like 4
Link to comment
Share on other sites

Yes, score is very definite on the winner and loser. But that is one data point. To extrapolate, one will add probability to it, which makes the data point fuzzy by a few points (say, one score), which is within our scoring margin. This means that we have been "lucky" falling within the right side of the fuzzy area, and statistically speaking, we are expected to fall on the wrong side of the fuzzy area in the future.

If this happens, S&P would be right. But if we have an uncanny ability to predominantly stay on the right side of the fuzzy area, then there is something to our game plan, style of play, etc. that is outside the norm, and hence be an exception to normative team play.

 

Link to comment
Share on other sites

We were up 19 on KSU, 21 on ou, and 13 on Baylor into the second half of the games before the gaps narrowed.  The computer doesn’t see this, it only gets fed the final stats. 

Some might put more weight on getting the 2 score leads, some may put more on giving them away. The computer doesn’t even see it. 

Link to comment
Share on other sites

2 hours ago, Vic Mackey said:

USC is 5-2. With one of their losses to us. They are not a bad team.

 

 

Just need them to beat ND at the end. Luckily the Pac 12 South is so bad, they should probably be 9-2 going into that home game. Last night was the perfect game for Texas. USC beat a ranked team while their offense looked like complete shit.

Link to comment
Share on other sites

2 hours ago, HornOnTheBayou said:

Not really when you're trying to predict how strong a team is. One point could change our perception of how strong the team is, but not how how strong they actually are. If Dicker misses that kick and we go on to lose to OU, it would definitely change all of our perceptions of how strong UT is. However, UT would still be the same team. One made or missed kick isn't going to change the strength of a team.

Some would argue that having a kicker who can make 50-yarders routinely but has trouble on do-or-die game-winners does, in fact, change how good a team is. The same principle applies to a player who fails to impress in practice or scrimmage but always seems to be money on third down late in the season. Statistics are helpful for modeling predictable behavior, but the problem lies in predicting outcomes in situations that incorporate a confounding variable.

Link to comment
Share on other sites

3 minutes ago, HtownHorn said:

Just need them to beat ND at the end. Luckily the Pac 12 South is so bad, they should probably be 9-2 going into that home game. Last night was the perfect game for Texas. USC beat a ranked team while their offense looked like complete shit.

They'll be underdogs in Utah this week.

Link to comment
Share on other sites

2 hours ago, Huckleberry said:

Yes, this is a huge factor. I haven't crunched their numbers enough but it's something I might disagree with if and when I do it. Drive efficiency is the point of football, it's essentially efficiency with each position. For example if you have an offense that gains 4 yards every play, so a mean of 4 and a standard deviation of 0, you will score a touchdown on every single drive where you don't commit a turnover. That is a perfect offense.

But I have a feeling it would score lower in S&P+ than an offense that averages 8 yards per play with a standard deviation of 16. But that "twice as explosive offense" has a probability of gaining a first down on three plays of about 70% compared to the 100% probability of the less explosive offense.

Basically. 

They do expected points based on yardline, down, and distance.  But looking just at yards gives an apt comparison.

Take these 12-yard series.

[4,4,4]

[-2,0,14]

YPP for both are 4.0 (the old S&P), YPP+ are respectively: 1.3 and 4.7.

So under S&P those give the same results, under S&P+ the second is more than 3 times better.  But any coach, player, or fan knows the first is better.

Link to comment
Share on other sites

1 minute ago, Pato del Muerto said:

Neither is better, they are the same. Both gain 12 yards in 3 plays.  One only looks better if you can only see one number at a time. 

No, the series of 4 yard gains is absolutely better since the entire point of this discussion is the predictive aspect of these ratings.

  • Like 1
Link to comment
Share on other sites

4 hours ago, Drew80 said:

I think it's a bit of both. We're winning, but not convincingly, and we keep finding ways to let our opponents stay in the game. Blowing a 21-point lead, missing two field goals, etc etc. We're much improved over last season, and with all our guys playing full throttle we look very good, but we're young and inconsistent.

That said, the model is clearly broken putting us at #43. Anyone who thinks North Texas would compete with us is stupid, but anyone who thinks we're a top 5 team is equally stupid.

North Texas would absolutely compete with us. And no we are not a Top 5 team. I doubt we are a Top 10 team. The score differential means something. We are likely a better Top 15 team than our scores versus MD, Tulsa, KSU and Baylor suggest. I think that is clear. 

This ranking is retarded, but so is #7 ahead of Georgia. We are a weird team, but I love our 6-game win streak and trajectory. 

Link to comment
Share on other sites

It's like people don't understand the philosophy of Herman. His offense isn't designed to score 50 points, it's designed to keep the ball by maximizing TOP and TO ratio. The 2 coaches that drove this home to him the most was Mack Brown and Urban Meyer. I bet it kills him he can't run the ball 50% of the plays for 200-250 yards per game.

This will likely be the least explosive highly successful offense Herman has moving forward as well.

Link to comment
Share on other sites

4 minutes ago, HtownHorn said:

It's like people don't understand the philosophy of Herman. His offense isn't designed to score 50 points, it's designed to keep the ball by maximizing TOP and TO ratio. The 2 coaches that drove this home to him the most was Mack Brown and Urban Meyer. I bet it kills him he can't run the ball 50% of the plays for 200-250 yards per game.

This will likely be the least explosive highly successful offense Herman has moving forward as well.

Yep. It's more widely dispersed, but it's still a cloud of dust. And when executed properly, I'll take it every time.

Link to comment
Share on other sites

S&P+ can't measure intangibles, and while I am not a believer in "grit" or "moxie," it's hard not to argue that some teams get better and others decline over the course of a season. Especially teams on the young side. Texas is young on O and was improving weekly with Sam, but SP doesn't give a shit about that. Word out of Auburn is that the team has lost faith in Stidham and by extension Malzahn, and they are getting worse as the season goes on, game by game. SP only measures what is sees on the field and won't work in a predicted decline for Auburn -- it will only report what Auburn has shown on the field. 

Connelly is getting a lot of shit for SP pegging App State in the top ten, but his counter to that was this. 


    @ 10 PSU    L45-38
    @ CHAR    W45-9
    vs USM    PPD
    vs WEBB    W72-7
    vs USA    W52-7
    @ ARKST    W35-9

A narrow OT loss on the road to a good to very good Penn State team and maulings of bad teams. That resume could just as easily belong to Michigan or Ohio State if they scheduled a bunch of patsies. One of those teams, Ark State, beat Tulsa by an almost identical score as our game -- 29-20 -- and App State worked them on the road. 

 

Link to comment
Share on other sites

8 hours ago, Vic Mackey said:

Who the fuck cares about this nerdy shit. We are 6-1 and are about to be ranked near the top 5 in the only polls that actually matter, not this junk.

Why even play games?

i guess “settling it on the field ..” is passé in the interwebs age: virtual> actual. Appearance >reality.

oh well

  • Like 1
Link to comment
Share on other sites

I don't know much about the S&P+ but I know several people that put a lot of stock in it.

I think our record is better than we are, and some of our playmakers are pulling us through. Discounting Maryland, we allowed Tulsa to fight back into a game they shouldn't have. Then we mauled USC and TCU because special teams and turnovers blew open close games. Kansas State clawed back into a game they should have been out of. Oklahoma same. Baylor same.

Our most dominant play was against Oklahoma, yet we still allowed them to come back from 3 scores down in the 4th quarter.

So yeah, we're having a charmed season. Sam avoiding sacks, some playmakers making things happen, and our turnover margin is a big part of it. I can see us dropping a few games. 

TCU and North Texas should not be ranked above us in any system.

Link to comment
Share on other sites

Having just one more point than the opponent is monumentally different than one fewer. The next 10 or 100 don’t change the outcome. People are not robots, and the importance of that one point and the accompanying victory can be lost on computers which don’t care, only calculate. 


It matters a great deal to who won that game but it has little, if any, predictive value as to how those teams will perform in the future.
Link to comment
Share on other sites

16 hours ago, AeroHorn said:

Yes, score is very definite on the winner and loser. But that is one data point. To extrapolate, one will add probability to it, which makes the data point fuzzy by a few points (say, one score), which is within our scoring margin. This means that we have been "lucky" falling within the right side of the fuzzy area, and statistically speaking, we are expected to fall on the wrong side of the fuzzy area in the future.

If this happens, S&P would be right. But if we have an uncanny ability to predominantly stay on the right side of the fuzzy area, then there is something to our game plan, style of play, etc. that is outside the norm, and hence be an exception to normative team play.

 

I know more about baseball than S&P+, but an analogy here could be batting average on balls in play (BABIP) - a hitter's batting average on playable batted balls (so, HR/SF/Ks factored out). Over several years and thousands of players, the norm for BABIP is right around .300. 

Now a guy may hit .350 for a month, or half the season...and the .350 represents hits he actually got, that counted. But if his BABIP is at .400 during this period, you can reasonably assume that his performance to date in unsustainable and that he has had some good fortune to get to that .350 mark. Expect regression.

Regarding the highlighted portion though - some guys have a skill that allows them to carry a higher-than-average BABIP and they may be able to sustain it. (Maybe they don't hit many grounders, they hit the ball consistently hard, etc.) 

Thing is - this all comes back to sample size. In baseball, you generally get an answer about whether what you're seeing is luck or skill, and it is all considered noise until there is a lot of data to support it. In 6 college football games, I'm not sure how you can get a very clear answer. 

Link to comment
Share on other sites

22 minutes ago, bigcigar said:

@Huckleberry so aggy is the best 2 loss team in the nation and should be ranked right below us?

That's where they fall in the current EWP ratings. In reality "best" team can be defined lots of different ways and they've also been fortunate to play a bunch of teams who can't throw the ball while fielding a defense strong against the run that is shitty against the pass.

  • Like 1
Link to comment
Share on other sites

Shit, Texas would be 8.5 point dogs to Appalachian State at home, with a 36.3% chance of winning the game.
 
Some should go on twitter and ask this douche if he'd bet his house on Appalachian State -8.5 at Texas.


He’s a douche because you disagree with his rating system?

His system doesn’t say Appalachian State would beat Texas by 9 or more 100% of the time. Hell, it says Texas would straight up win 36% or the time. So even taking the Appalachian State moneyline, he would expect to lose his house over 1/3 of the time.
Link to comment
Share on other sites

Any ranking that has us at 46 is garbage.

But I’d agree we’re not back. We’re a lot better than we’ve been, but until we start putting away lesser teams in the first half and getting all our backups in the game, we won’t be back imo. I don’t think anyone would be surprised if we lost any of the remaining games save Kansas.

Link to comment
Share on other sites

49 minutes ago, Huckleberry said:

That's where they fall in the current EWP ratings. In reality "best" team can be defined lots of different ways and they've also been fortunate to play a bunch of teams who can't throw the ball while fielding a defense strong against the run that is shitty against the pass.

Yep.....the SEC quarterbacking group (outside of Tua) as a whole has been much worse than expected. Still think the ags have 3 more losses in them somewhere. 

Link to comment
Share on other sites

These models recall the Popism that, “a little learning is a dangerous thing.” There are, all told, literally innumerable variables that affect the final outcome of a football game. For someone interested in whether the variables he has identified have some effect on the outcome, developing a model that guesses right 50ish percent of the time is positive but far from definitive evidence. There is 50% difference between blind luck and certainty in predicting winners, and the S&P+ eliminates less than 20% of that uncertainty. The rest is not known, but likely includes things we have trouble measuring and predicting but which our brain assesses as part of its gestalt. The predictor says OSU is more likely to win than we, but our ‘irrational, emotional’ analysis disagrees. Certainly emotion can play into our disagreement, but there is far more. S&P+ is like 40 time. Recruits with faster 40 times, ceteris paribus, are preferable to those who are slower. Good recruiters don’t always prefer the faster 40 recruit in absolutely all circumstances. A slavish deference to S&P+ as better than educated observer opinion would be analogous to ignoring off-field issues and his team’s offensive line in RB recruiting.

Link to comment
Share on other sites

Remember when sweater vest used to win all those games by slim margins up in Ohio? A W is a W, championships count the same.
I’d love to Alabama everyone, but we’re learning to win, and after the last 8 years, that matters more to me than anything else.
That's the problem, they're measuring everyone on the Bama scale.

But to be honest, they're in a category of their own most years for the better part of a decade.
Link to comment
Share on other sites

Wasn't there a tOSU team back in the early 00s that barely won a bunch of games but then made the NC game? Was that the year Miami beat them? I wonder what their S&P+ looked like throughout that year.
Bama measuring stick didn't exist back then..
Link to comment
Share on other sites

42 minutes ago, Magus Ossis said:

There are, all told, literally innumerable variables that affect the final outcome of a football game. For someone interested in whether the variables he has identified have some effect on the outcome, developing a model that guesses right 50ish percent of the time is positive but far from definitive evidence. There is 50% difference between blind luck and certainty in predicting winners, and the S&P+ eliminates less than 20% of that uncertainty. 

Exactly - the article explaining S&P+ states that the ratings predict outcomes against the spread at a slightly higher 50% rate. These seem to be models designed specifically for betting.   

Link to comment
Share on other sites

5 minutes ago, Hank Chinaski said:

Exactly - the article explaining S&P+ states that the ratings predict outcomes against the spread at a slightly higher 50% rate. These seem to be models designed specifically for betting.   

And if you had taken Baylor and the points you would have cleaned up. Maybe because Sam went down, or maybe not. 

Link to comment
Share on other sites

14 hours ago, MaybeACoordinator said:

S&P+ can't measure intangibles, and while I am not a believer in "grit" or "moxie," it's hard not to argue that some teams get better and others decline over the course of a season. Especially teams on the young side. Texas is young on O and was improving weekly with Sam, but SP doesn't give a shit about that. Word out of Auburn is that the team has lost faith in Stidham and by extension Malzahn, and they are getting worse as the season goes on, game by game. SP only measures what is sees on the field and won't work in a predicted decline for Auburn -- it will only report what Auburn has shown on the field. 

Connelly is getting a lot of shit for SP pegging App State in the top ten, but his counter to that was this. 


    @ 10 PSU    L45-38
    @ CHAR    W45-9
    vs USM    PPD
    vs WEBB    W72-7
    vs USA    W52-7
    @ ARKST    W35-9

A narrow OT loss on the road to a good to very good Penn State team and maulings of bad teams. That resume could just as easily belong to Michigan or Ohio State if they scheduled a bunch of patsies. One of those teams, Ark State, beat Tulsa by an almost identical score as our game -- 29-20 -- and App State worked them on the road. 

 

So Tulsa is a common opponent of Texas and ark st. App st beating ark st means something in relation to Texas?  How much weight is placed on third degree relationships like this?

Link to comment
Share on other sites

6 minutes ago, Pato del Muerto said:

So Tulsa is a common opponent of Texas and ark st. App st beating ark st means something in relation to Texas?  How much weight is placed on third degree relationships like this?

Honestly, I don't know. I listen to Connelly's podcasts and I am learning about S&P Plus little by little, via his explanations every week, rather than trying to absorb it all through reading about it. My thinking is that is less about the transitive property than it is about dominating teams with secondary stats.

Link to comment
Share on other sites

Clock management is an example of a difference between teams that tend somewhat to win versus lose close games. Two teams that play fairly evenly for 57 minutes can have the game decided by which coach is better at being in the position to win by last-second FG. My guess is that if you look at KSU over the course of Snyder's tenure, they over-performed the S&P+ W/L predictions. Similarly, teams with QBs that their teams love and trust are more likely to see an extra ounce of 4th quarter effort. I would expect that Texas didn't merely win more often with better QBs, but over-performed S&P+ W/L projections in games that VY and Colt started and finished, versus those with QBs that haven't been 'emotionally' said to have 'it.' In at least these senses, winning may actually be predictive of winning, independent of noticeably different aggregate numbers in explosive plays, strength of schedule, MOV, etc.

Link to comment
Share on other sites

19 minutes ago, Magus Ossis said:

 My guess is that if you look at KSU over the course of Snyder's tenure, they over-performed the S&P+ W/L predictions. 

You would be correct. 

https://www.footballstudyhall.com/2017/6/1/15723948/college-football-coaching-underachievers-overachievers

Check out who landed number three on the overachiever list.  

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...