S&P+ rankings: Overperformance or overcorrection?

[Drew 2331]

Explain to me how OU has a 81% chance to beat Texas in a rematch?

If you want to say you think OU wins the rematch, that's cool.  Even give them more than a 50% chance for whatever reason.  But 81%?!?

Texas dominated that game on both lines that entire game, and the score has made people think it was closer than it was.  Simple case of a young team not keeping their foot on the pedal/or on their necks to close it out.

 

 

[satyanash 14276]

5 minutes ago, Drew said:

Explain to me how OU has a 81% chance to beat Texas in a rematch?

If you want to say you think OU wins the rematch, that's cool.  Even give them more than a 50% chance for whatever reason.  But 81%?!?

Texas dominated that game on both lines that entire game, and the score has made people think it was closer than it was.  Simple case of a young team not keeping their foot on the pedal/or on their necks to close it out.

Because in S&P+'s eyes our win in Dallas was fluky. blOU outgained us both through the air and on the ground, and was significantly more explosive in doing so. We won because we were +3 in turnovers, which on most days simply isn't going to happen. This is how Bill explains it:

Quote

"Another unlikely result from this week: Texas over Oklahoma. The Horns had a 9% post-game win expectancy against the Sooners, getting out-gained by 2.5 yards per play and allowing a 61% success rate. But that one’s at least easier to understand: turnovers, turnovers, turnovers.

Texas had a plus-3 turnover margin when stats suggest about a plus-1.4 margin was much more likely. Turnovers are worth about five points in field position lost and gained, so the God of randomness gifted the Horns about eight points in a three-point win."

 

[Drew 2331]

How much of the air/ground yardage, and how many of those explosive plays were in the 4th quarter during their comeback?

 

Sometimes the numbers don't show everything.

[Huckleberry 32685]

Forced fumbles and interceptions aren't random, and if he actually thinks he can predict them based on yardage and rate stats then that's pretty questionable.

[TheYoungHorn 57]

3 minutes ago, Huckleberry said:

Forced fumbles and interceptions aren't random, and if he actually thinks he can predict them based on yardage and rate stats then that's pretty questionable.

No. I think his turnover factor is only influenced by sacks and, I believe, passes defended, which he claims are the only quantifiable factor in kept data that correlates to causing turnovers. But you might know better than me how he comes to that number

Edited October 24, 2018 by TheYoungHorn

[satyanash 14276]

6 minutes ago, TheYoungHorn said:

No. I think his turnover factor is only influenced by sacks and, I believe, passes defended, which he claims are the only quantifiable factor in kept data that correlates to causing turnovers. But you might know better than me how he comes to that number

I think the two factors that go into it are forced fumbles and pass breakups. It doesn't look at how many actual turnovers you get because that figure can rarely be predicted (turnovers are highly dependent on luck), so it's considered an anomaly when your actual turnover margin is out of line with what's expected based on the above two factors.

[dcar00 14023]

This is all about beating the shit out of crappy to midling teams.  we aren't doing that yet but we sure did beat the shit out of OU for 3 quarters.  If we are healthy when we see them again in Dec, we will do it again.

[rickyspub 801]

24 minutes ago, dcar00 said:

This is all about beating the shit out of crappy to midling teams.  we aren't doing that yet but we sure did beat the shit out of OU for 3 quarters.  If we are healthy when we see them again in Dec, we will do it again.

But it's not. As aggy has only close wins over middling teams and a blowout over one of Connelly's lowest rated teams. They have two losses to top teams, one close and one a blowout. Their 7th game doesn't even count in Connelly's ratings. And they are ranked over 20 places higher than us. I have a feeling if you changed our Tulsa win to a 30 point blowout we magically jump 25 spots in his rankings. 

Edited October 24, 2018 by rickyspub

[Magus Ossis 2773]

Aggy will have to make room in their trophy case for their Texas State S&P+ October Championship trophy. I’ll just go cry in my beer and try to console myself by reflecting on our actual wins.

[Pig Bellmont 8679]

1 hour ago, satyanash said:

Because in S&P+'s eyes our win in Dallas was fluky. blOU outgained us both through the air and on the ground, and was significantly more explosive in doing so. We won because we were +3 in turnovers, which on most days simply isn't going to happen. This is how Bill explains it:

 

It’s a good thing we actually play the games. Also, Texas dominated that game. I know we almost blew it at the end when coaches got conservative, but that easily could have been a 2-3 TD blow out. It’s a shame their shitty defense won’t allow us to play them again 

[DanRydell 799]

6 hours ago, Pig Bellmont said:

Click baiting bullshit. When Texas wins the BigXII and OU loses to either WVU or Tech this guy really hopes he’ll be tweeted about so he can claim relevance 

He's been on Football Outsiders and had his own Football Study Hall vertical on SBNation for near a decade now. He's also had a podcast for a few years now. He's doing fine for himself.

[JBJ 3647]

Connelly's actually a level-headed guy and smartly downplays the model in his actual writing (not the headlines). He'll tell you why the model says what it says and the reasons it might be wrong or where it might be right.

Edited October 24, 2018 by JBJ

[Machinator 32097]

Yeah, anyone who reads Connelly's work (especially his preseason team previews) knows that he isn't afraid to challenge what his models are telling him.

[HtownHorn 14446]

2 hours ago, satyanash said:

Because in S&P+'s eyes our win in Dallas was fluky. blOU outgained us both through the air and on the ground, and was significantly more explosive in doing so. We won because we were +3 in turnovers, which on most days simply isn't going to happen. This is how Bill explains it:

 

Shit, another issue with his fucking model. The last TO came on the last play of the game when OU was throwing the ball around for shits and grins.

[Gaffords 755]

It should really factor in Sam coming into a rematch with a confidence level that will be difficult to shake. While Murry's will be looking for a reason to fold.  I think this will be the biggest factor in a rematch.  

This OU team looks like a Mack Brown led Texas. It was looking for a reason to collapse against us.  The half time locker room meltdown point to severe issues that did not get fixed overnight. It's signs of a cancer in the senior class that didn't sign up for Riley as a head coach and don't buy into his system.  They will fold if Texas comes out with a chip and hits them in the mouth (again).

[taybo20 2815]

S&P+ has pokes winning 33-26

[satyanash 14276]

3 hours ago, taybo20 said:

S&P+ has pokes winning 33-26

S&P+ has Okie State pegged as the toughest team left on our schedule. If we can get out of Stillwater with a win, we'll be in great position heading into the WVU/TTU/ISU stretch.

[satyanash 14276]

 

[satyanash 14276]

 

[TheYoungHorn 57]

To update the discussion earlier on computer models that account for recency/trends, this is from collegefootballpoll: 

Quote

OKLAHOMA STATE by 1.04 over Texas (-3) - Saturday, October 27.
The Cowboys (4-3, 1-3 Big 12) are uncharacteristically reeling from 3 losses in their last 4 games. Conversely, the Longhorns (6-1, 4-0 Big 12) are soaring with 6 straight wins after the inexplicable season-opening loss to Maryland. The computer trend line favors Texas by 5.

 

[bigcigar 1161]

Looks like the model was wrong. Should have had OSU by 50.

[MaybeACoordinator 8349]

We are (less than) what Connelly thought we were. 

[satyanash 14276]

so yeah... Connolly might know what he's talking about 😥

[TheYoungHorn 57]

Systems that take into account every single play (minus truly garbage time snaps) are as unbiased of an arbiter you will find in this sport. 

Those systems, which do have their faults and blind spots (many of which we have addressed in this thread), didn't believe Texas was anywhere close to a top-10 team, and tonight is just another data point to confirm it. 

Before you argue that some (insert officiating call or no call here) would have changed the outcome, and Texas sneaks out of Stillwater with another close win, realize that Texas doesn't play consistent enough football to be a top 10 team---something that many here, and Connelly have teased out ad naseum. 

The Herman coached team narrative continues to be oddly true. 

 

[satyanash 14276]

 

[GabrielsHorn 2142]

Should we point to S&P+ whenever Texas knocks off OU? Or USC? or TCU?

Seems like the whole SBNation gang is down to do pop shots whenever that happens, but how about their predictions for App State/Georgia Southern, Miami/BC, etc etc? 

[TheYoungHorn 57]

Just now, GabrielsHorn said:

Should we point to S&P+ whenever Texas knocks off OU? Or USC? or TCU?

Seems like the whole SBNation gang is down to do pop shots whenever that happens, but how about their predictions for App State/Georgia Southern, Miami/BC, etc etc? 

Which is why it's crazy to determine any model, or set of model's, legitimacy based on a single game (the ultimate of small sample size). 

What some people here were arguing is that the model is fundamentally flawed because it gave Texas such poor chances to win against teams like OSU, WVU, ISU and TTU, when in fact I think any reasonable fan that has watched Texas play every quarter of every game this year would say, "Yeah, I have concerns that if they continue to play the way they have all year, and not just in a half of TCU, USC and OU, that they will struggle to win all of those games." 

We can do this every week and cherry-pick results where these models look prophetic or aloof, but it doesn't address the underlying issues of why Texas is not getting love in these models. 

They are not getting love because when you step back and look at the pure inputs they don't do things as well, consistently as teams that win all (or most) of their games. 

 

[bigcigar 1161]

S&P+ was technically wrong this week. Had OSU covering 3.5

[GabrielsHorn 2142]

Except in a model this wide, there will always be teams that exceed their predictions based on what the selected inputs are. The obvious issue with this specific model is it only looks at the end of game stats and not necessarily the game situations (the obvious example is OU). 

It’s a good model, but seeing people take it as gospel is what rubs a lot of people the wrong way. It’s decent as a predictor, but note that Bill uses the opening lines and not those closer towards the end of the week. Last I checked on Todd Beck’s prediction tracker, it’s behind FPI and Saragon. I think it gets its infallibility from Bill having a great podcast (PAPN) and a lot of the blogs in CFB coming from SBNations platform which obviously would push it. 

[TheYoungHorn 57]

Just curious, I looked up past Herman teams to see how they fared in the eyes of Connelly's model: 

2016---Houston finished 9-4, and was ranked 39th in final S&P+

2015--Houston finished 13-1, and was ranked 46th in final  S&P+

[TheYoungHorn 57]

3 minutes ago, GabrielsHorn said:

Except in a model this wide, there will always be teams that exceed their predictions based on what the selected inputs are. The obvious issue with this specific model is it only looks at the end of game stats and not necessarily the game situations (the obvious example is OU). 

It’s a good model, but seeing people take it as gospel is what rubs a lot of people the wrong way. It’s decent as a predictor, but note that Bill uses the opening lines and not those closer towards the end of the week. Last I checked on Todd Beck’s prediction tracker, it’s behind FPI and Saragon. I think it gets its infallibility from Bill having a great podcast (PAPN) and a lot of the blogs in CFB coming from SBNations platform which obviously would push it. 

Not sure where, or who, finds it infallible.

Anyone treating any model as gospel is silly, especially when it's creator will publicly and consistently admit it's shortcomings and tease out why those might apply in particular cases. 

I think it gets the credit it deserves, because Connelly (and the properties he publishes, or partners with) provide it as accessible and open information. 

The only thing that is funny is when the team you follow happens to be one of the outliers in a model, and fans start to feel like some computer model is an enemy that shouldn't just be addressed, but attacked. 

 

 

 

[satyanash 14276]

I'll be honest, I was really hoping we'd pull off the comeback so I could tweet this @ Bill

[MaybeACoordinator 8349]

We probably would have fallen in S&P+ even with a win, thanks to USC and TCU.

[satyanash 14276]

Updated S&P+ outlook for Texas. Right now, the model projects us to go 8-4.

[bigcigar 1161]

10 minutes ago, satyanash said:

Updated S&P+ outlook for Texas. Right now, the model projects us to go 8-4.

That seems pretty accurate

[Phoenix 65]

So people are using one data point to justify ranking Texas so low? The S&P+ is very far off from all the other ranking systems, and that's because it uses asinine statistics to come up with the rating. The only thing that matters in football is the SCORE. You can gain all the yards you want, in whatever manner you want, but it doesn't matter if it doesn't put points on the board.

[satyanash 14276]

51 minutes ago, Phoenix said:

So people are using one data point to justify ranking Texas so low? The S&P+ is very far off from all the other ranking systems, and that's because it uses asinine statistics to come up with the rating. The only thing that matters in football is the SCORE. You can gain all the yards you want, in whatever manner you want, but it doesn't matter if it doesn't put points on the board. 

Everything you just said has been addressed earlier in the thread. At least read those posts.

S&P+ is predictive. So yeah, it cares about more than the score. Any competently-designed predictive mode would. It routinely beats Vegas and it doesn't do so using "asinine" statistics.

[I Plead Da Fif 240]

7 hours ago, bigcigar said:

That seems pretty accurate

And would fucking suck

[Phoenix 65]

If a team gets 600 yards but only scores 20 points, but the team they play against gets 400 yards and scores 23, does it matter who had more yards? Any ranking system that had Texas in the 40's after being 6-1 at that point in time, with a tough SOS in tow, immediately disqualifies itself.

[TXSooner518 2432]

If a team gets 600 yards but only scores 20 points, but the team they play against gets 400 yards and scores 23, does it matter who had more yards? Any ranking system that had Texas in the 40's after being 6-1 at that point in time, with a tough SOS in tow, immediately disqualifies itself.

I’m not sure if it is your reading comprehension or just your general ability to understand that is the problem here.

Determining who won a football game looks at the score of that game. Determining who is more likely to win a future football game looks at how good the teams actually are, which is reliant on more than just the scores of the games.

The Giants beat the Patriots in the 2007 Super Bowl. Yet if they had played again the next week, the Pats would have been double digit favorites. Figure out why and maybe you will get it.

[DanRydell 799]

If a team gets 600 yards but only scores 20 points, but the team they play against gets 400 yards and scores 23, does it matter who had more yards?

If you’re trying to project how those teams will perform moving forward, yes.

[Mikey4 2476]

If you’re trying to project how those teams will perform moving forward, yes.

If a team can rush down the field but has no red zone offense, then they will struggle to win games notwithstanding the nice rushing stats.

[Dark Horse 2257]

43 minutes ago, Mikey4 said:

 

 

If a team can rush down the field but has no red zone offense, then they will struggle to win games notwithstanding the nice rushing stats.

 

Scoring efficiency handles that issue. Points per possession. 

[Mikey4 2476]

Scoring efficiency handles that issue. Points per possession. 

Agreed, scoring points is what matters. Not how many yards you get.

[TXSooner518 2432]

13 minutes ago, Mikey4 said:

53 minutes ago, Dark Horse said:

Scoring efficiency handles that issue. Points per possession. 

Agreed, scoring points is what matters. Not how many yards you get.

Over time, yards generally correlate to points. If one team averages 6.0 ypp, and another team averages 3.5 ypp, the first team will average a much higher scoring rate, and thus projects to win more games. 

[MaybeACoordinator 8349]

Connelly nailed it again this week. S&P+ had WVU -3 (given home field -- it'd have been 6 at a neutral site).  

[TheYoungHorn 57]

2 hours ago, MaybeACoordinator said:

Connelly nailed it again this week. S&P+ had WVU -3 (given home field -- it'd have been 6 at a neutral site).  

S&P sucks. WVU didn't even cover that

[satyanash 14276]

Another year, another blOU team in the driver's seat to win the Big 12. Fan-fucking-tastic.

 

[Huckleberry 32685]

lulz

As if turning the ball over isn't part of the game.

When Texas isn't giving up scoring drives, nobody can touch them.

[satyanash 14276]

S&P+ now predicts us to win 7.8 games.