Jump to content

S&P+ rankings: Overperformance or overcorrection?


satyanash

Recommended Posts

11 hours ago, satyanash said:

An interesting note: The 2016 team that lost to Kansas and got Strong fired finished the year ranked #37 in S&P+ despite their dismal 5-7 record.

Most people agree about the Kansas schools’ and Baylor’s likely poor final standings and about OU’s likely finishing 8-1 (or close to it). The more controversial and useful predictions are who else goes to the CCG and what is the pecking order for the other bowls. Texas at 5 and TCU at 7 is an interesting result. It will be interesting to see how S&P+ does over the back half of the season and revisit this list on T+2.

Edited by Magus Ossis
Link to comment
Share on other sites

39 minutes ago, Magus Ossis said:

Most people agree about the Kansas schools’ and Baylor’s likely poor final standings and about OU’s likely finishing 8-1 (or close to it). The more controversial and useful predictions are who else goes to the CCG and what is the pecking order for the other bowls. Texas at 5 and TCU at 7 is an interesting result. It will be interesting to see how S&P+ does over the back half of the season and revisit this list on T+2.

Boldest prediction there is Texas at 6-3 by far, and I am no kool aid sipper. That equates to 1-3 vs OSU, Tech, ISU, and WVU assuming a win against KU. We couldn't even pull that off last year or in 2016.

Link to comment
Share on other sites

Boldest prediction there is Texas at 6-3 by far, and I am no kool aid sipper. That equates to 1-3 vs OSU, Tech, ISU, and WVU assuming a win against KU. We couldn't even pull that off last year or in 2016.


S&P+ actually projects 7-2 as the most likely Big 12 record for Texas (34% probability) with 6-3 second most likely (31% probability).

https://docs.google.com/spreadsheets/d/e/2PACX-1vQ2e9xV7-ClihFVJ3kla0ZDxzFCQ7-WXvQRur-nK6gOzo333PqSetw52kEGgbXKb6viGZSbYuJugvRR/pubhtml#

The author of that tweet simply gave each team a win or loss based on the projected winner of each game, which ignores that having a 5% chance to win and a 48% chance to win are extremely different.
  • Like 1
Link to comment
Share on other sites

24 minutes ago, Pato del Muerto said:

7-2 with a win over wvu gets us in the ccg irrespective of what the rest of the league does to each other.  

I think. 

Nope.  Tech winning out would definitely bump us.  If we finish 7-2 tech could lose another as long as they beat us and ISU winning out would bump us out. 

Link to comment
Share on other sites

S&P predicts Texas is more likely to lose than win 3 of our remaining games. The 7-2 outcome assumes we luck into a win in one of those games despite being the weaker team. The problem with aggregate records like that is that they don’t assign wins in specific contests; they merely look for the hump of the win probability curve in total. To sort teams including tie-breakers requires knowing who beat whom. If you want granular S&P+ predictions, now you get us at 6-3 instead of 7-2.

Link to comment
Share on other sites

3 hours ago, Treefidy said:

Nope.  Tech winning out would definitely bump us.  If we finish 7-2 tech could lose another as long as they beat us and ISU winning out would bump us out. 

I haven't looked at this close enough so correct me if I'm wrong. 

1) If Tech wins out that means OU would take another loss since they still have to play each other. It also means Iowa State would take another loss since they still have to play each other. That means OU would finish with 2 losses. It also means ISU would finish with 3 losses. 

Assuming Tech wins out would mean that we wouldn't be bumped by ISU or OU.

I'm not sure how the multi-team tiebreaker works in the Big 12 but we own head to head with OU. So if it isn't a multi-team tie with 2 losses we would get in over OU. Actually, we would own head to head wins over all the Big 12 teams that could possibly finish with 2 losses in this scenario (OU, TCU, Baylor, and WV). Again, I'm not sure how the multi-team tiebreaker works. But I would be pissed if we owned head to head against every team with 2 losses and didn't make the Big 12 championship game. 

2) I think there's a very good chance we make the Big 12 championship game with a win against WV even if we finish with 2 losses. Almost a certainty.

I say that because does anyone see Baylor, ISU, or TCU running the table? All 3 of these teams are going to drop another game and finish with 3 losses IMO. Those 3 are eliminated and we already own head to head against 2 of those teams. 

Tech still has @ Iowa State, OU, Texas, @ K-State, @ Baylor. That's a really difficult finish and I bet they still have at least 2 losses on the schedule. 

I think this is a 3 team race between OU, Texas, and WV. If we beat WV ( and end up with 2 losses in the Big 12) we would own head to head and there would have to be some real fuckery for a multi-team tie to take place. 

 

 

 

 

 

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

6 hours ago, texasstrong12 said:

I haven't looked at this close enough so correct me if I'm wrong. 

1) If Tech wins out that means OU would take another loss since they still have to play each other. It also means Iowa State would take another loss since they still have to play each other. That means OU would finish with 2 losses. It also means ISU would finish with 3 losses. 

Assuming Tech wins out would mean that we wouldn't be bumped by ISU or OU.

I'm not sure how the multi-team tiebreaker works in the Big 12 but we own head to head with OU. So if it isn't a multi-team tie with 2 losses we would get in over OU. Actually, we would own head to head wins over all the Big 12 teams that could possibly finish with 2 losses in this scenario (OU, TCU, Baylor, and WV). Again, I'm not sure how the multi-team tiebreaker works. But I would be pissed if we owned head to head against every team with 2 losses and didn't make the Big 12 championship game. 

2) I think there's a very good chance we make the Big 12 championship game with a win against WV even if we finish with 2 losses. Almost a certainty.

I say that because does anyone see Baylor, ISU, or TCU running the table? All 3 of these teams are going to drop another game and finish with 3 losses IMO. Those 3 are eliminated and we already own head to head against 2 of those teams. 

Tech still has @ Iowa State, OU, Texas, @ K-State, @ Baylor. That's a really difficult finish and I bet they still have at least 2 losses on the schedule. 

I think this is a 3 team race between OU, Texas, and WV. If we beat WV ( and end up with 2 losses in the Big 12) we would own head to head and there would have to be some real fuckery for a multi-team tie to take place. 

 

 

 

 

 

That's right, forgot about Tech hasn't played OU.  So if we beat WV and finish 7-2 only ISU bumps us by winning out and Tech bumps us by winning out except to OU who would have to also win out.  

If ISU beats Tech, their remaining schedule basically only has us left.  Baylor, Kansas, KSU, incarnate word.  That seems like the more likely, they are favored in all but our game  

if tech beats ISU they have remaining Baylor, Kansas, KSU, OU and us.  Tech is favored in all but the ISU and OU games. 

Predictor has us favored in all but tech game  

 

Link to comment
Share on other sites

As far as conference tiebreaker:

a 2 team tiebreaker is head to head.

a multi-team tiebreaker is determined:

1. head to head in mini round robin format.  

2. Compare remaining conference results,  (this one is the most confusing with 2 side notes)

3. Scoring differential between the tied teams.  

4. Draw

Link to comment
Share on other sites

7 minutes ago, satyanash said:

Updated S&P+ rankings are out for the week: https://www.sbnation.com/college-football/2018/10/21/18005470/ncaa-football-rankings-2018-week-9

We jumped 5 spots to #38 over the bye week. Next week we face Oklahoma State, sitting at #22. Highest-ranked team in the Big 12 is blOU at #3.

LSU is still below Georgia?  Makes sense

Link to comment
Share on other sites

1 hour ago, EuroHorn said:

LSU is still below Georgia?  Makes sense

why can people not grasp that it is not ranking the past results, but predicting future performance?  LSU should be ranked ahead of UGA and Texas should be ranked ahead of OU. Nonetheless, if they were playing next week on a neutral field, UGA and OU would be the favorites.

  • Like 1
Link to comment
Share on other sites

Texas' win probability distribution for the rest of the regular season.
texas%20-%20sp+%20-%20red-green.png
S&P+ has us at almost exactly between 8 and 9 wins.


Barely a 1-out-of-3 chance to win in Stillwater? Granted it’s never an easy task there, but this year’s OSU team pales in comparison to years’ past.
Link to comment
Share on other sites

3 hours ago, bigcigar said:

And 53% for the season last year...so didn’t even cover the vig.

I’m not saying it’s completely worthless, but it’s flawed and the Texas/OU comparison exposes its flaws.

53% does in fact cover the vig.

Consistent 55% is outstanding. 60% is other-worldly.

Link to comment
Share on other sites

Exactly.  Just because you beat someone doesn't mean than you can beat them.


Just because you beat someone doesn’t mean you are or were more than 50% to beat them. The last big poker tourney I played, I was 97% to win a hand when the money went in. I lost. My hand will still be 97% to win in the poker S&P rankings.
Link to comment
Share on other sites

19 hours ago, TXSooner518 said:

why can people not grasp that it is not ranking the past results, but predicting future performance?  LSU should be ranked ahead of UGA and Texas should be ranked ahead of OU. Nonetheless, if they were playing next week on a neutral field, UGA and OU would be the favorites.

Because people don't know how to read. 

  • Like 1
Link to comment
Share on other sites

16 hours ago, bigcigar said:

You’re right...barely. That’s a .3% return. Or with 100 $11 bets, $13 profit.

How are you performing this season while betting every single game? 

Predicting college football games is incredibly hard, and while the model is far from perfect, his five factors have done a good job of helping better predict who will win games based on the factors that determine successful football teams from publicly available box-score and play-by-play data. 

If he was any better at it, I doubt he would be publicly sharing the data with the whole world, as he would be busy using his black box to make a fortune in Vegas. 

Can we all just admit that Texas has played a lot of bad football this season even while winning almost all of their games, and that his system ranks all of those bad stretches and plays equally as the good football we have also played?

We can disagree on if the bad stretches of limping to close results against bad teams is some super secret, MENSA-Jedi strategy to sandbag the S&P+ rankings so analytics-obsessed opponents will under rank Texas and Longhorn fans will have a new enemy to direct their hate toward, and not indicative of the team they are (or can be) as a result, but it's easy to see why this particular Texas team does so poorly in this rankings. 

  • Like 1
Link to comment
Share on other sites

I understand why the rankings might not love Texas. I also understand how they’re predictive and why OU is ahead of them.

But to anyone who has seen OSU play and seen Texas play this year...I really think the inputs need some tweaking given that they come up with that result. We might lose, sure. But they’ve clearly looked like the inferior team to anyone with eyes.

Link to comment
Share on other sites

6 minutes ago, formermav43 said:

I understand why the rankings might not love Texas. I also understand how they’re predictive and why OU is ahead of them.

But to anyone who has seen OSU play and seen Texas play this year...I really think the inputs need some tweaking given that they come up with that result. We might lose, sure. But they’ve clearly looked like the inferior team to anyone with eyes.

While I agree that it feels wrong to have OSU nearly a 2-1 favorite, I think that there are a few factors that make S&P less predictive here.  For one, Texas seems to be playing better as the season moves on, while OSU seems to be playing worse, and fans naturally bake that in to their evaluations. If last week, Texas lost to Maryland while OSU boatraced Boise, Texas fans would be very pessimistic about this game.  I don't know that S&P factors in recency.

Also, I don't think that S&P factors in that Iowa State's top QB and their awesome RB were both out for the game against OSU, yet ISU scored 48. Fans know this and that game looks worse to them then it would to a computer.

Those factors show why the Texas-OSU line differs pretty wildly from S&P.  Based only on S&P you would expect OSU to be favored by 5-6, yet Texas is favored, despite unclear health for Ehlinger.

Link to comment
Share on other sites

5 minutes ago, TXSooner518 said:

While I agree that it feels wrong to have OSU nearly a 2-1 favorite, I think that there are a few factors that make S&P less predictive here.  For one, Texas seems to be playing better as the season moves on, while OSU seems to be playing worse, and fans naturally bake that in to their evaluations. If last week, Texas lost to Maryland while OSU boatraced Boise, Texas fans would be very pessimistic about this game.  I don't know that S&P factors in recency.

Also, I don't think that S&P factors in that Iowa State's top QB and their awesome RB were both out for the game against OSU, yet ISU scored 48. Fans know this and that game looks worse to them then it would to a computer.

Those factors show why the Texas-OSU line differs pretty wildly from S&P.  Based only on S&P you would expect OSU to be favored by 5-6, yet Texas is favored, despite unclear health for Ehlinger.

Exactly, there is no "trending" model that I know of baked into this particular model. 

Some of the other computer models do have that, for example the Cosgrove model at collegefootballpoll.com who is .544 ATS this year. And his model has Texas as -1.04 this weekend in Stillwater. 

For reference, his model also has Texas ranked at #12 (http://www.collegefootballpoll.com/rankings/)

 

Edited by TheYoungHorn
Link to comment
Share on other sites

53 minutes ago, TheYoungHorn said:

Exactly, there is no "trending" model that I know of baked into this particular model. 

Some of the other computer models do have that, for example the Cosgrove model at collegefootballpoll.com who is .544 ATS this year. And his model has Texas as -1.04 this weekend in Stillwater. 

For reference, his model also has Texas ranked at #12 (http://www.collegefootballpoll.com/rankings/)

 

His picks are showing OSU at -1.04 which seems a little odd when his rankings have Texas #12 and OSU #45.

 

It would be interesting to take those predictive rankings and pick the top ten or so games that had the biggest discrepancy when compared to the betting line and see how that has performed over the year. 

Edited by NeverMarryAStripper
Link to comment
Share on other sites

Resume S&P+ rankings through week 8: https://www.sbnation.com/college-football/2018/8/29/17795292/college-football-2018-strength-of-schedule-week-9-rankings

It looks like this is Bill Connolly's attempt to create a metric that is both forward and backward looking. Combines S&P+ with strength of schedule.

 

  • Like 1
Link to comment
Share on other sites

Blowouts are the toughest thing for any system to deal with. The issue I have with some of his stuff is he has the data to take a better approach, the problem is correlating the data to an adjusted margin in each game.

What I mean is this: Alabama being up 49-7 on Ole Miss at halftime and then winning the game 62-7 means that it goes into his system as a 55-point win. It was worse than that. Same as being up 45-16 in the third quarter and then shutting it down against A&M to win 45-23.

A team that is up, say, 38-10 after three quarters and wins the game 38-24 should have a much better adjusted margin than a team that was up 31-24 after three quarters, sweated out the fourth, then returned a fumble for a touchdown with 30 seconds to go. He mocks the "use your eyes" arguments but he has actual data that he can use to approximate that aspect of using your eyes to watch a game.

Because he doesn't do that, Alabama's win over Arkansas State only looks 24 points worse than Appalachian State's win over the Red Wolves. In reality Alabama was up 40-0 at halftime so that 57-7 final score ends up giving App State too much credit in his system. I'm not ripping on Connelly here because this is an issue with nearly every margin of victory system, but it also exists in his Résumé S&P+.

Edited by Huckleberry
  • Like 5
Link to comment
Share on other sites

I would say use multiple computer ratings if you want to have a good comparison between two teams. S&P+ is obviously biased against a team like ours and perhaps there are a few models that overrate us. Though I imagine we underperform the broad majority of computer models just because we don't blow teams out. I think our S&P+ spot would be 15-20 places higher if we just beat Tulsa by 40. In the real world, I am not sure what that would tell us differently about this team moving ahead.

Link to comment
Share on other sites

Blowouts are the toughest thing for any system to deal with. The issue I have with some of his stuff is he has the data to take a better approach, the problem is correlating the data to an adjusted margin in each game.
What I mean is this: Alabama being up 49-7 on Ole Miss at halftime and then winning the game 62-7 means that it goes into his system as a 55-point win. It was worse than that. Same as being up 45-16 in the third quarter and then shutting it down against A&M to win 45-23.
A team that is up, say, 38-10 after three quarters and wins the game 38-24 should have a much better adjusted margin than a team that was up 31-24 after three quarters, sweated out the fourth, then returned a fumble for a touchdown with 30 seconds to go. He mocks the "use your eyes" arguments but he has actual data that he can use to approximate that aspect of using your eyes to watch a game.
Because he doesn't do that, Alabama's win over Arkansas State only looks 24 points worse than Appalachian State's win over the Red Wolves. In reality Alabama was up 40-0 at halftime so that 57-7 final score ends up giving App State too much credit in his system. I'm not ripping on Connelly here because this is an issue with nearly every margin of victory system, but it also exists in his Résumé S&P+.


I don’t think regular S&P+ even knows whether you won or lost a game, let alone by how much. It simply knows how you played within that game. And it ignores all data while the game is in garbage time. So for your Bama-Ole Miss example, it would never even know the second half of that game occurred.
Link to comment
Share on other sites

Bill Connolly: We might already know 3 Playoff teams. Who do the numbers see at No. 4?

Of the remaining one-loss P5 teams, we have the lowest chance (3%) of winning out the rest of the regular season and the highest chance (82%) of ending up with 3 or more losses.

Quote

With so many teams losing over the last couple of weeks, Oklahoma is back in prime Playoff position despite the loss to Texas. The Sooners are up to third in S&P+ (first on offense, naturally) and have at least a 69% win probability in each remaining game. On a neutral field in the Big 12 title game, they’d have an 81% chance of winning a rematch against Texas and a 74% chance of beating WVU. Winning out is hard, but the Sooners have the best shot among the pool of one-loss teams.

 

Link to comment
Share on other sites

8 minutes ago, satyanash said:

Bill Connolly: We might already know 3 Playoff teams. Who do the numbers see at No. 4?

Of the remaining one-loss P5 teams, we have the lowest chance (3%) of winning out the rest of the regular season and the highest chance (82%) of ending up with 3 or more losses.

 

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 

  • Like 1
Link to comment
Share on other sites

1 hour ago, GabrielsHorn said:

Is it a surprise that a model that doesn’t like us expects us to crash and burn?

It shouldn’t be. We look beatable most weeks. 7 games in and I’m still not really sure what to think about this team. It can play really well then look like dog shit within minutes.

Link to comment
Share on other sites

1 hour ago, bigcigar said:

It shouldn’t be. We look beatable most weeks. 7 games in and I’m still not really sure what to think about this team. It can play really well then look like dog shit within minutes.

Everybody but Alabama looks beatable. Pitt struck the fear of god in Notre Dame until ND finally took the lead with 5:43 left in the 4th. A 24-16 win over a 3-5 Ball State at home isn't what I would call pretty but everybody still loves Notre Dame. Clemson's butt cheeks were puckered in the 4th quarter against Syracuse. LSU had only a 3 point lead on Louisiana Tech early in the 4th, Michigan trailed Northwestern 17-7 before winning 20-17, Oklahoma needed OT to beat Army, Ohio State just got waxed by Purdue...

Why are we the only team in college football that isn't allowed to look human?

Link to comment
Share on other sites

8 minutes ago, Texas25 said:

Everybody but Alabama looks beatable. Pitt struck the fear of god in Notre Dame until ND finally took the lead with 5:43 left in the 4th. A 24-16 win over a 3-5 Ball State at home isn't what I would call pretty but everybody still loves Notre Dame. Clemson's butt cheeks were puckered in the 4th quarter against Syracuse. LSU had only a 3 point lead on Louisiana Tech early in the 4th, Michigan trailed Northwestern 17-7 before winning 20-17, Oklahoma needed OT to beat Army, Ohio State just got waxed by Purdue...

Why are we the only team in college football that isn't allowed to look human?

Because it has happened in all but 2 games with one of those being a loss to Maryland.

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...