ESPN Pimping SEC

[satyanash 14129]

4 minutes ago, Ricky Butler said:

What are you saying?

That it's not just ESPN bias. S&P+, another computer prediction model, is unaffiliated with ESPN yet is also down on us. The true reason probably lies elsewhere.

[Richter 2195]

2 minutes ago, Ricky Butler said:

Looking at these FPI rankings, I see two losses and fodder for A&M bloggers to claim their 4-5 loss Ags are better than us. Just the right amount of salt for schadenfreude stew.

The model heavily weights the quality of losses and Aggy’s losses are like *the* best losses.

[sushihorn 3215]

3 minutes ago, satyanash said:

That it's not just ESPN bias. S&P+, another computer prediction model, is unaffiliated with ESPN yet is also down on us. The true reason probably lies elsewhere.

One from a guy who basically admits his model is useless for projecting Tom Herman teams.  The other blatantly biased towards the SEC and promoted by a network with $100 million invested in the conference brand.  So yes, there are Separate But Equal reasons for the bias.

[Ricky Butler 2636]

 

1 hour ago, satyanash said:

That it's not just ESPN bias. S&P+, another computer prediction model, is unaffiliated with ESPN yet is also down on us. The true reason probably lies elsewhere.

Can you make a clear statement on what you think? I'm not antagonizing. Do you think we're in for a down year because of the models?

[satyanash 14129]

18 minutes ago, Ricky Butler said:

Can you make a clear statement on what you think? I'm not antagonizing. Do you think we're in for a down year because of the models?

I think with what we're losing, next year could be the 2007 to our 2006. Lot of youth in starting positions and trying to adjust. The 2007 team also had to replace a lot of experienced starters while keeping the same QB, and although we still went 10-3 it definitely felt like it wasn't as strong a season overall. Obviously it was a prelude to a very strong 2008 (2020 equivalent) team, but I'm not sure where that leaves us going into next season. S&P+ seems to acknowledge this as well by being bearish on us, although maybe to too great a degree, and it looks like FPI does so as well. The models may be reinforcing my prognostication, not creating them.

Edited February 20, 2019 by satyanash

[Ricky Butler 2636]

17 minutes ago, satyanash said:

I think with what we're losing, next year could be the 2007 to our 2006. Lot of youth in starting positions and trying to adjust. The 2007 team also had to replace a lot of experienced starters while keeping the same QB, and although we still went 10-3 it definitely felt like it wasn't as strong a season overall. Obviously it was a prelude to a very strong 2008 (2020 equivalent) team, but I'm not sure where that leaves us going into next season. S&P+ seems to acknowledge this as well by being bearish on us, although maybe to too great a degree, and it looks like FPI does so as well. The models may be reinforcing my prognostication, not creating them.

I guess it's all about expectations because that doesn't seem like a terrible outcome for next year, especially if one of our wins is LSU, OU, or a good bowl opponent. I think recruiting keeps humming along in that scenario.

[sushihorn 3215]

49 minutes ago, satyanash said:

I think with what we're losing, next year could be the 2007 to our 2006. Lot of youth in starting positions and trying to adjust. The 2007 team also had to replace a lot of experienced starters while keeping the same QB, and although we still went 10-3 it definitely felt like it wasn't as strong a season overall. Obviously it was a prelude to a very strong 2008 (2020 equivalent) team, but I'm not sure where that leaves us going into next season. S&P+ seems to acknowledge this as well by being bearish on us, although maybe to too great a degree, and it looks like FPI does so as well. The models may be reinforcing my prognostication, not creating them.

Sorry dude.  10-3 is nothing close to what that fool Connelly is projecting.  He has Texas as the #35 team in the country,  That's basically at the bottom of the "others receiving votes" or slightly below that.  With the difficulty of the 2019 schedule, the Horns would have to be 8-5 or maybe 7-6 with some good victories to sink that low.  This FPI isn't much better with Texas at #26.

The reason why they're wrong should be obvious to anyone who examines their models.  They both use a big dose of past years' performance as inputs.  Four years back in the case of FPI and FIVE years for S&P+.  Anybody think it's realistic to project Tom Herman's 2019 Longhorns based on Charlie Strong's record?  Of course not; that's idiocy but it's precisely what both these models do.  Connelly will still have Millstone Charlie dragging Texas down when he does his 2021 projections.  Tons of blind spots in these models.  The worst unstated assumption may be that coaching doesn't matter.

To understand why Baylor is in the same region as Texas in these projections, just remember that the later Art Briles years are part of their calculation.  Nevermind that they've changed coaches twice and turned over the entire team.  The methodology really only works for programs that experience no significant coaching change during the look back window.

Edited February 20, 2019 by sushihorn

[sushihorn 3215]

***Rant Alert***

The first thing you need to understand is that FPI and S&P+ along with many other statistical models don't necessarily take account of what actually happened.  They take data inputs than simulate what the model says SHOULD have happened.  IOW, they value simulation over reality. 

How else can one explain the utter stupidity of their rankings at the end of the 2018 season, after all the results were in?  At that point, S&P+ ranked Texas as the #35 team in the country, 21 spots behind Washington with the same record and a weaker SOS, 20 spots behind 9-4 Pedo State and 15 spots behind 7-6 S Carolina.  It's a model so grotesque that it ranked 5-7 Ole Miss one spot below 10-4 Texas.  To top it off, the same "model" rated 2017 Texas higher than 2018 Texas.  Those fools need to pull their heads out of their spreadsheets and look around at reality.

FPI, while still moronic looks like Deep Blue compared to the dumpster fire of S&P+.  They ranked Texas as the #19 team at the end of the 2018 season.  How on earth they think the #19 team could split with their #5 team and utterly dominate their #3 team is a mystery but whatever.  One thing that computers can't do is recognize that the Texas team that destroyed Georgia in the Sugar Bowl was not at all the same team that lost to Maryland to open the season.  Humans can see that, which is why the coaches and sports writers both had Texas at #9.

[rickyspub 801]

8 minutes ago, sushihorn said:

***Rant Alert***

The first thing you need to understand is that FPI and S&P+ along with many other statistical models don't necessarily take account of what actually happened.  They take data inputs than simulate what the model says SHOULD have happened.  IOW, they value simulation over reality. 

How else can one explain the utter stupidity of their rankings at the end of the 2018 season, after all the results were in?  At that point, S&P+ ranked Texas as the #35 team in the country, 21 spots behind Washington with the same record and a weaker SOS, 20 spots behind 9-4 Pedo State and 15 spots behind 7-6 S Carolina.  It's a model so grotesque that it ranked 5-7 Ole Miss one spot below 10-4 Texas.  To top it off, the same "model" rated 2017 Texas higher than 2018 Texas.  Those fools need to pull their heads out of their spreadsheets and look around at reality.

FPI, while still moronic looks like Deep Blue compared to the dumpster fire of S&P+.  They ranked Texas as the #19 team at the end of the 2018 season.  How on earth they think the #19 team could split with their #5 team and utterly dominate their #3 team is a mystery but whatever.  One thing that computers can't do is recognize that the Texas team that destroyed Georgia in the Sugar Bowl was not at all the same team that lost to Maryland to open the season.  Humans can see that, which is why the coaches and sports writers both had Texas at #9.

And at the same time both act like aggy won't miss a beat. Sure we may have lost more starters in terms of sheer numbers, but as HTown notes they lost substantially more in terms of sheer production, but they are somehow almost a Top 10 team and we are in the 20s or 30s.

[sushihorn 3215]

13 hours ago, satyanash said:

That it's not just ESPN bias. S&P+, another computer prediction model, is unaffiliated with ESPN yet is also down on us. The true reason probably lies elsewhere.

In statistical terms there is a strong recency bias in both models.  That works if there is no dramatic change in the program over the measured period.  It's pretty much a guaranteed fail when there is either a major upgrade or major downgrade in coaching.  In theory the recruiting component should capture changes in talent on the field.  There is nothing to factor in changes in talent on the sideline (and coaches' box).

For any team switching its HC from Charlie Strong to Tom Herman that assumption is a really bad one.

Edited February 20, 2019 by sushihorn

[Huckleberry 32142]

11 hours ago, satyanash said:

I think with what we're losing, next year could be the 2007 to our 2006. Lot of youth in starting positions and trying to adjust. The 2007 team also had to replace a lot of experienced starters while keeping the same QB, and although we still went 10-3 it definitely felt like it wasn't as strong a season overall. Obviously it was a prelude to a very strong 2008 (2020 equivalent) team, but I'm not sure where that leaves us going into next season. S&P+ seems to acknowledge this as well by being bearish on us, although maybe to too great a degree, and it looks like FPI does so as well. The models may be reinforcing my prognostication, not creating them.

The 2007 team was pretty much exactly as good as the 2006 team. 

[markstanco 12424]

One from a guy who basically admits his model is useless for projecting Tom Herman teams. 

This is the 3rd or 4th time I've heard this but haven't seen what it means. So, what does it mean?

[Fico 1724]

11 hours ago, satyanash said:

I think with what we're losing, next year could be the 2007 to our 2006. Lot of youth in starting positions and trying to adjust. The 2007 team also had to replace a lot of experienced starters while keeping the same QB, and although we still went 10-3 it definitely felt like it wasn't as strong a season overall. Obviously it was a prelude to a very strong 2008 (2020 equivalent) team, but I'm not sure where that leaves us going into next season. S&P+ seems to acknowledge this as well by being bearish on us, although maybe to too great a degree, and it looks like FPI does so as well. The models may be reinforcing my prognostication, not creating them.

Except nobody had high expectations for 2008, that is why it was awesome.  Our JR QB took a massive leap to Heisman caliber play that year, and our young DBs turned out to be players across the board. Hell, many people were still clamoring for "Chiles to start."

Sam's Sophomore year was significantly better than Colt's, and honestly if you have good to elite QB play, you are more than likely going to be a good to elite team.  Everyone knows 2020 is supposed to be a good year, but if Sam can continue to improve on last season, we are going to be an extremely tough team to beat, and will likely dominate several lesser opponents. 

[sushihorn 3215]

13 minutes ago, markstanco said:

This is the 3rd or 4th time I've heard this but haven't seen what it means. So, what does it mean?

Tom Herman is the coach that has the largest positive win deviation from what the S&P+ model says should happen.  His results are approximately 3 std deviations above the norm - which is Mensa indeed.  Put another way, Herman is the coach who S&P+ tends to underestimate most badly, which of course we could easily infer from Connelly ranking Herman's Longhorns #35.

https://www.footballstudyhall.com/2019/2/12/18221578/college-football-coaching-underachievers-overachievers-2019

Coaching overachievers and underachievers

Coach	Years	Diff Wins/Year	Percentile
Tom Herman	4	1.49	99.8%

[satyanash 14129]

41 minutes ago, Huckleberry said:

The 2007 team was pretty much exactly as good as the 2006 team. 

Surprised you think this. That 2007 team faced a much weaker schedule.

The 2006 team was on track to win the Big 12, with only one loss against #1 Ohio State, before Colt got hurt which ruined everything. The 2007 squad got blown out at home by a 5-7 K-State team, and also suffered losses to the two other strong Big 12 South squads in blOU and A&M. They also benefited by avoiding the two best Big 12 North teams that year, Missouri and Kansas. In fact, I'm pretty sure our best win in 2007 was against ASU in the bowl game.

[Huckleberry 32142]

It's not just me. Most people think that.

2006 Texas - #16 EWP, #27 Massey, #19 Sagarin, #13 AP, #13 Coaches
2007 Texas - #15 EWP, #19 Massey, #12 Sagarin, #10t AP, #10t Coaches

[Burt Macklin 18960]

14 hours ago, satyanash said:

That it's not just ESPN bias. S&P+, another computer prediction model, is unaffiliated with ESPN yet is also down on us. The true reason probably lies elsewhere.

Yeah. The FPI is strictly a model, so it has nothing to do with ESPN bias. Although both ESPN and FPI seem to overrate quality losses and blowing out horrible opponents, which does favor one conference over the others, but I don’t attribute that to intentional personal/organizational bias.

12 hours ago, satyanash said:

I think with what we're losing, next year could be the 2007 to our 2006. Lot of youth in starting positions and trying to adjust. The 2007 team also had to replace a lot of experienced starters while keeping the same QB, and although we still went 10-3 it definitely felt like it wasn't as strong a season overall. Obviously it was a prelude to a very strong 2008 (2020 equivalent) team, but I'm not sure where that leaves us going into next season. S&P+ seems to acknowledge this as well by being bearish on us, although maybe to too great a degree, and it looks like FPI does so as well. The models may be reinforcing my prognostication, not creating them.

The big differences between 2019 and 2007 is that the disparity in young talent replacing the previous year’s starters is massive for 2019, whereas it wasn’t in 2007. 

That’s what makes projecting the 2019 team so tough. Programs almost never experience such a radical change in coaching and talent as ours has with going from Charlie to Herman and also going from the 2015 class (way worse than its ranking dueattrition, best talent leaving in 2017), 2016 (attrition killed this class) and 2017 (#25) to what we’ve brought in in 2018 and 2019. 

Having said that, it wouldn’t be that hard for us to get back to 10 wins again even if we’re not much, of at all, better than last year.

3 hours ago, rickyspub said:

And at the same time both act like aggy won't miss a beat. Sure we may have lost more starters in terms of sheer numbers, but as HTown notes they lost substantially more in terms of sheer production, but they are somehow almost a Top 10 team and we are in the 20s or 30s.

But A&M was ranked way higher than us last year, so their drop is from a higher place in the rankings. Like I said above, the systems definitely overrate SEC steams due to giving too much credit for blowing out bad teams, or more specifically by not devaluing ranking factors like play efficiency and explosiveness against terrible opponents, like the 1AA teams SEC schools schedule.

Edited February 20, 2019 by Burt Macklin

[rickyspub 801]

20 minutes ago, Burt Macklin said:

But A&M was ranked way higher than us last year, so their drop is from a higher place in the rankings. Like I said above, the systems definitely overrate SEC steams due to giving too much credit for blowing out bad teams, or more specifically by not devaluing ranking factors like play efficiency and explosiveness against terrible opponents, like the 1AA teams SEC schools schedule.

I didn't check to see where they were last year, but I don't think they dropped as much as we did. I pretty much ignore FPI, but the S&P stuff has a whole thread devoted to it and I agree that S&P overweighs the results of games against G5 teams (and FCS games, though I am not sure if those are included in S&P). As far as returning player models go, S&P's model is completely generic and ignores context. Aggy losing Williams (who had 70% of their rushing yards and 1400+ yards more than the next RB) is treated the same as us losing Watson (who had only 40% of our total yards and less than 80 yds more than Ingram).  

[Burt Macklin 18960]

1 minute ago, rickyspub said:

I didn't check to see where they were last year, but I don't think they dropped as much as we did. I pretty much ignore FPI, but the S&P stuff has a whole thread devoted to it and I agree that S&P overweighs the results of games against G5 teams (and FCS games, though I am not sure if those are included in S&P). As far as returning player models go, S&P's model is completely generic and ignores context. Aggy losing Williams (who had 70% of their rushing yards and 1400+ yards more than the next RB) is treated the same as us losing Watson (who had only 40% of our total yards and less than 80 yds more than Ingram).  

Yeah. That’ll always be an issue with these pre-season model rankings. They account for quality of talent lost and quality of talent replacing with a very broad brush, because it would be close to impossible to give a specific ranking for each position on each team.

[rickyspub 801]

1 minute ago, Burt Macklin said:

Yeah. That’ll always be an issue with these pre-season model rankings. They account for quality of talent lost and quality of talent replacing with a very broad brush, because it would be close to impossible to give a specific ranking for each position on each team.

Honestly, with the amount of data Connelly claims he is crunching to get his generic roster loss factors, he should be able spin out something a little less stupid. He has to grab all the data regardless to keep his factors up to date year on year. Either he doesn't have time to grab the data early enough to put out his click-bait pre-season rankings or his generic factors are bullshit.

[Burt Macklin 18960]

4 minutes ago, rickyspub said:

Honestly, with the amount of data Connelly claims he is crunching to get his generic roster loss factors, he should be able spin out something a little less stupid. He has to grab all the data regardless to keep his factors up to date year on year. Either he doesn't have time to grab the data early enough to put out his click-bait pre-season rankings or his generic factors are bullshit.

I mean what data would you want him to use to determine the level of quality of individual players coming and going? If there were reliable advanved stats/metrics for individual players, that could work, but that doesn’t exist for college football. Maybe he could do something like subtracting adjusted line yards from a RB, but that would get iffy and only helps for one position.  

Weighing production stats and starts for OL is about the only way to do it when applying this metric to so many teams, which is why these kind of models are inherently flawed.

Edited February 20, 2019 by Burt Macklin

[satyanash 14129]

21 minutes ago, rickyspub said:

Honestly, with the amount of data Connelly claims he is crunching to get his generic roster loss factors, he should be able spin out something a little less stupid. He has to grab all the data regardless to keep his factors up to date year on year. Either he doesn't have time to grab the data early enough to put out his click-bait pre-season rankings or his generic factors are bullshit.

Can't really use production stats to gauge individual player quality. For example, Wheeler led the team with eight tackles and a sack against WVU yet it was easily his worst ever game as a Longhorn.

[Ricky Butler 2636]

23 minutes ago, Burt Macklin said:

Yeah. That’ll always be an issue with these pre-season model rankings. They account for quality of talent lost and quality of talent replacing with a very broad brush, because it would be close to impossible to give a specific ranking for each position on each team.

I think this the big thing for me. How good the players are is like the most important thing in sports. If you can't do that really accurately, you're going to have some big misses. It's like trying to do an NBA model where the teams only play 14 games and there is no standard level of play from year to year to compare performance. If someone like Connelly could do that, then his name would be HaralaBill Connelly and he'd be working in a front office.

[Richard Kimball 1491]

12 hours ago, Ricky Butler said:

I guess it's all about expectations because that doesn't seem like a terrible outcome for next year, especially if one of our wins is LSU, OU, or a good bowl opponent. I think recruiting keeps humming along in that scenario.

Big challenge in figuring out Texas for this year is the number of one score games from last year, against both good and bad teams. There were seven wins by a score or less and three losses by a score or less. So, if we figure all those games went down to the final possession, it's a possible record of anywhere from 13-1 to 4-10, depending on a couple of bounces in each game. 

Good sign is the Horns have gotten more physical, and physicality travels well. Herman may make some calls I wouldn't make, but I usually understand the logic. There are far fewer WTF moments than there were under Strong. Ehlinger is looking good, and there are only a few returning strong quarterbacks. Thinking Brewer and Purdy are solid. Anyone else? Four new coaches, but the Horns were 3-1 against the teams that replaced coaches.  I THINK the Horns turned the corner. 

Weird note: Herman has been at UT for two years. There are now only three Big 12 coaches that have been in the league longer than him; Patterson, Campbell and Gundy. 

[rickyspub 801]

1 minute ago, satyanash said:

Can't really use production stats to gauge individual player quality. For example, Wheeler led the team with eight tackles and a sack against WVU yet it was easily his worst ever game as a Longhorn.

Then you can't use it as any sort of metric then. I would think knowing Wheeler accounted for 30% of the team's tackles could be useful in noting how his departure might affect us moving forward. Connelly instead knocks us some set amount of points regardless of whether he was our leading tackler or a JAG starter. Basically Wheeler hurts our 'returning starter' factor in S&P the same as it would have when Derrick Johnson left. That is a pretty brain dead factor to me.

[Ricky Butler 2636]

Just now, rickyspub said:

Then you can't use it as any sort of metric then. I would think knowing Wheeler accounted for 30% of the team's tackles could be useful in noting how his departure might affect us moving forward. Connelly instead knocks us some set amount of points regardless of whether he was our leading tackler or a JAG starter. Basically Wheeler hurts our 'returning starter' factor in S&P the same as it would have when Derrick Johnson left. That is a pretty brain dead factor to me.

I would be curious if he has more accurately predicted Missouri's performance in the past seeing as how familiar he is with their personnel and recruiting.

[rickyspub 801]

24 minutes ago, Burt Macklin said:

I mean what data would you want him to use to determine the level of quality of individual players coming and going? If there were reliable advanved stats/metrics for individual players, that could work, but that doesn’t exist for college football. Maybe he could do something like subtracting adjusted line yards from a RB, but that would get iffy and only helps for one position.  

Weighing production stats and starts for OL is about the only way to do it when applying this metric to so many teams, which is why these kind of models are inherently flawed.

At the very least you could measure percentage production loss at a position instead of acting like us losing Watson, who was barely our leading rusher, is the same as aggy losing not only their leading rusher but a guy that gained 1700 yds with no one of note behind him. aggy may very well repeat that result with whoever replaces Williams, but I would imagine it is much more likely that Ingram easily replaces Watson's production and puts up a season more like Williams's.

Connelly claims the numbers he uses for scoring player loss is based on prior performance. Did he randomly select a few teams to see how they did year over year? That seems like a recipe for shitty data. If he pulled from all teams, then he has the data to look deeper.

[Burt Macklin 18960]

5 minutes ago, rickyspub said:

At the very least you could measure percentage production loss at a position instead of acting like us losing Watson, who was barely our leading rusher, is the same as aggy losing not only their leading rusher but a guy that gained 1700 yds with no one of note behind him. aggy may very well repeat that result with whoever replaces Williams, but I would imagine it is much more likely that Ingram easily replaces Watson's production and puts up a season more like Williams's.

Connelly claims the numbers he uses for scoring player loss is based on prior performance. Did he randomly select a few teams to see how they did year over year? That seems like a recipe for shitty data. If he pulled from all teams, then he has the data to look deeper.

I think you may be confused on Connelly’s system here. That article in the S&P thread with husband preseason rankings explains his system. My understanding is he  does measure percentage production lost. He uses specific production data like yards rushing, receiving, tackles, etc and multiplies those by weighted percentages to determine the production lost value. So basically, he already does what you said. He doesn’t just say both teams lost their leading rusher, so it’s the same value.

 

[satyanash 14129]

3 minutes ago, Burt Macklin said:

I think you may be confused on Connelly’s system here. That article in the S&P thread with husband preseason rankings explains his system. My understanding is he  does measure percentage production lost. He uses specific production data like yards rushing, receiving, tackles, etc and multiplies those by weighted percentages to determine the production lost value. So basically, he already does what you said. He doesn’t just say both teams lost their leading rusher, so it’s the same value.

 

He also calculates a Havoc rate for each individual defensive player, measured as the amount of disruption they cause to the opposing offense (passes defensed and forced fumbles are two of the stats that go into it, among others). Kris Boyd led the team in Havoc Rate which is why he counts him leaving as a very big loss.

[rickyspub 801]

13 minutes ago, Burt Macklin said:

I think you may be confused on Connelly’s system here. That article in the S&P thread with husband preseason rankings explains his system. My understanding is he  does measure percentage production lost. He uses specific production data like yards rushing, receiving, tackles, etc and multiplies those by weighted percentages to determine the production lost value. So basically, he already does what you said. He doesn’t just say both teams lost their leading rusher, so it’s the same value.

 

Got it. I thought his weights were used to account for the fact that some positions don't necessarily have specific data to record. So a lost lineman is worth say .222 while a lost QB is worth more at .444 or something like that. 

[HtownHorn 14285]

6 hours ago, sushihorn said:

***Rant Alert***

The first thing you need to understand is that FPI and S&P+ along with many other statistical models don't necessarily take account of what actually happened.  They take data inputs than simulate what the model says SHOULD have happened.  IOW, they value simulation over reality. 

How else can one explain the utter stupidity of their rankings at the end of the 2018 season, after all the results were in?  At that point, S&P+ ranked Texas as the #35 team in the country, 21 spots behind Washington with the same record and a weaker SOS, 20 spots behind 9-4 Pedo State and 15 spots behind 7-6 S Carolina.  It's a model so grotesque that it ranked 5-7 Ole Miss one spot below 10-4 Texas.  To top it off, the same "model" rated 2017 Texas higher than 2018 Texas.  Those fools need to pull their heads out of their spreadsheets and look around at reality.

FPI, while still moronic looks like Deep Blue compared to the dumpster fire of S&P+.  They ranked Texas as the #19 team at the end of the 2018 season.  How on earth they think the #19 team could split with their #5 team and utterly dominate their #3 team is a mystery but whatever.  One thing that computers can't do is recognize that the Texas team that destroyed Georgia in the Sugar Bowl was not at all the same team that lost to Maryland to open the season.  Humans can see that, which is why the coaches and sports writers both had Texas at #9.

It also doesn't take into account that Texas was up usually by 17+ points in every game they played except a few.

Up 21-0 against Tulsa

Up 19-0 against KSU

Up 45-24 against OU

Up 23-10 against BU

Up 27-10 and 34-17 against TT

Up 24-3 against ISU

Up 21-0 against KU

Up 28-7 against UGA

Texas had trouble finishing off those teams, but dominated each of those games for long stretches.

[Burt Macklin 18960]

 

Then you can't use it as any sort of metric then. I would think knowing Wheeler accounted for 30% of the team's tackles could be useful in noting how his departure might affect us moving forward. Connelly instead knocks us some set amount of points regardless of whether he was our leading tackler or a JAG starter. Basically Wheeler hurts our 'returning starter' factor in S&P the same as it would have when Derrick Johnson left. That is a pretty brain dead factor to me.

This is the whole point, though. These metrics are general approximations and will always have flaws but it’s as close as you can get without individually scouting each player and assigning them a grade, like PFF (whose college ratings suck because they don’t spend as much time on them as the NFL ones) and even then you’d struggle to account for talent of the player replacing him. Recruiting rankings would be the closest but still problematic.

this is why when people get all worked up about UT’s rating, it’s silly. Like I said above, the disparity in past performance due to coaching and change in talent level of the program explain why UT is a huge outlier and likely won’t be properly ranked by these metrics.

 

I would be curious if he has more accurately predicted Missouri's performance in the past seeing as how familiar he is with their personnel and recruiting.

He is creating a formula, so he’s not inputting any extra info for Missouri, even if he knows more about their roster than other schools.

A nationwide formula will never be exact.

 

 

[Burt Macklin 18960]

 

Got it. I thought his weights were used to account for the fact that some positions don't necessarily have specific data to record. So a lost lineman is worth say .222 while a lost QB is worth more at .444 or something like that. 

Offensive Linemen is based on starts (I think), but I believe every other position uses specific stats that are then multiplied by a weighting formula he created.

[Tex Long 4264]

Why is this thread? WTF do y'all (any of y'all) give a shit about Connelly and his predictions? Do you gamble, is he an indicator of anything worthwhile? 

[LTtxfan 48688]

37 minutes ago, HtownHorn said:

It also doesn't take into account that Texas was up usually by 17+ points in every game they played except a few.

Up 21-0 against Tulsa

Up 19-0 against KSU

Up 45-24 against OU

Up 23-10 against BU

Up 27-10 and 34-17 against TT

Up 24-3 against ISU

Up 21-0 against KU

Up 28-7 against UGA

Texas had trouble finishing off those teams, but dominated each of those games for long stretches.

Interesting Point.... thx

[Ricky Butler 2636]

42 minutes ago, Burt Macklin said:

He is creating a formula, so he’s not inputting any extra info for Missouri, even if he knows more about their roster than other schools.

A nationwide formula will never be exact.

 

 

I understand that. My thought is if he's ever tried to compete against S&P, on a micro level, using better inputs. 

[rickyspub 801]

48 minutes ago, Tex Long said:

Why is this thread? WTF do y'all (any of y'all) give a shit about Connelly and his predictions? Do you gamble, is he an indicator of anything worthwhile? 

Honestly, only I gave a shit because Connelly is a pussy in his articles and little bitch on Twitter, especially towards Texas.

I am not a bettor, but it sounds like some years his predictions would pay off...well, if you bet the same amount on every single college football game. Some one else said he used Connelly's picks in a pool of select games and lost his shirt. So in aggregate I guess he is better than throwing darts, but if you are betting on individual games you are probably better off researching the details yourself and deciding which lines are most favorable.

[TornACL 7652]

1 hour ago, HtownHorn said:

It also doesn't take into account that Texas was up usually by 17+ points in every game they played except a few.

Up 21-0 against Tulsa

Up 19-0 against KSU

Up 45-24 against OU

Up 23-10 against BU

Up 27-10 and 34-17 against TT

Up 24-3 against ISU

Up 21-0 against KU

Up 28-7 against UGA

Texas had trouble finishing off those teams, but dominated each of those games for long stretches.

I had not realized this until you posted them all in one place. We were also up 37-14 against USC. It's pretty insane that each and every one of those games got "interesting" with the exception of USC, and I guess Iowa State.

I knew we had trouble closing things out, but holy fuck.  

[ztejas 42744]

2 hours ago, satyanash said:

He also calculates a Havoc rate for each individual defensive player, measured as the amount of disruption they cause to the opposing offense (passes defensed and forced fumbles are two of the stats that go into it, among others). Kris Boyd led the team in Havoc Rate which is why he counts him leaving as a very big loss.

He certainly caused a lot of havoc in remotes almost thrown through TV sets.

[satyanash 14129]

30 minutes ago, TornACL said:

I had not realized this until you posted them all in one place. We were also up 37-14 against USC. It's pretty insane that each and every one of those games got "interesting" with the exception of USC, and I guess Iowa State.

I knew we had trouble closing things out, but holy fuck.  

We were outscored 148-66 in the 4th quarter this season. Worst in the country, I believe. You'd never guess we had a 10-win season by looking at that one stat 😂

Edited February 20, 2019 by satyanash

[Beau Vine 53625]

19 hours ago, satyanash said:

That it's not just ESPN bias. S&P+, another computer prediction model, is unaffiliated with ESPN yet is also down on us. The true reason probably lies elsewhere.

Nah, ESPN bias has infiltrated S&P+.  They even took two of their letters from ESPN.  Heelllllllooooooo?

[HtownHorn 14285]

41 minutes ago, TornACL said:

I had not realized this until you posted them all in one place. We were also up 37-14 against USC. It's pretty insane that each and every one of those games got "interesting" with the exception of USC, and I guess Iowa State.

I knew we had trouble closing things out, but holy fuck.  

I didn't include USC because that game was over after the FG block, nor did I include TCU as Texas put that game away in the 4Q, but yes both fit nicely into the lead by 17+ narrative. The only games Texas lost were games they never went up by 17+ points, which underscores the 4Q deficit mentioned earlier.

[satyanash 14129]

4 hours ago, Huckleberry said:

It's not just me. Most people think that.

2006 Texas - #16 EWP, #27 Massey, #19 Sagarin, #13 AP, #13 Coaches
2007 Texas - #15 EWP, #19 Massey, #12 Sagarin, #10t AP, #10t Coaches

The seasons were equal sure, but only because of Colt's injury which knocked us out of the Big 12 title race by costing us the last two games. A healthy 2006 squad was significantly better than the healthy 2007 one.

[Valmy77 12519]

1 hour ago, satyanash said:

The seasons were equal sure, but only because of Colt's injury which knocked us out of the Big 12 title race by costing us the last two games. A healthy 2006 squad was significantly better than the healthy 2007 one.

And why did Colt get injured? Because our offensive line was bad and couldn't even move Sam Houston State. 

[Valmy77 12519]

1 hour ago, satyanash said:

We were outscored 148-66 in the 4th quarter this season. Worst in the country, I believe. You'd never guess we had a 10-win season by looking at that one stat 😂

Yeah and you are claiming we are going to regress from that...

[satyanash 14129]

And why did Colt get injured? Because our offensive line was bad and couldn't even move Sam Houston State. 

Lol he was injured on a bloody touchdown run, not because of anything our O-line did. The 2006 line had Sendlein, Studdard, and Consensus All-American Justin Blalock; it was better than it's 2007 successor.

 

 

 

[satyanash 14129]

FPI gives Clemson a staggering 83% chance of making the CFP again this year. Alabama is at 71%.

The Big 12: Oklahoma is your only hope

Sorry, Longhorns fans: This is not your season for the playoff. Despite hype building behind quarterback Sam Ehlinger and coach Tom Herman, the Longhorns simply need to replace too much talent to be a realistic contender in 2019. With just eight returning starters, Texas is unlikely to challenge Oklahoma the way it did in 2018.

Our Playoff Predictor gives Texas less than a 1% chance to reach the playoff this season. Which is also what it gives every other team in the Big 12 not named Oklahoma. The conference's hopes are pinned yet again on the Sooners, who have a 35 percent chance to return to college football's playoff for the third consecutive season. With Kyler Murray moving on to the NFL, that might feel high, but remember that coach Lincoln Riley and the rest of the team in Norman faced the same situation a year ago after Baker Mayfield departed. FPI does give Oklahoma credit for incoming transfer QB Jalen Hurts.

[ztejas 42744]

6 minutes ago, satyanash said:

Sorry, Longhorns fans: This is not your season for the playoff.

Well shit let's just not play the season then. 

Oh wait, I have a better idea, let's beat OU's ass twice and if we don't make the playoff they don't either. 

[Jersey Man10 3470]

Less than 1% chance huh?

[HtownHorn 14285]

Wait, OU has to replace Murray, Brown, Anderson, Meier, Evans, Powers, Samia, Ford, and Seibert on offense while also suffering massive hemorrhaging along the DL with defections, and Texas has too much talent to replace to challenge OU. 

Edited April 17, 2019 by HtownHorn