Jump to content

S&P+ rankings: Overperformance or overcorrection?

satyanash

By satyanash
in Football

 Share

Recommended Posts

  • Replies 1.1k
  • Created
  • Last Reply

Top Posters In This Topic

Popular Days

Top Posters In This Topic

Popular Days

Popular Posts

Dahobbs
Dahobbs

Fozzz is a fucking idiot. 

Magus Ossis
Magus Ossis

A model that drops the winner and sustains the loser of a game needs adjusting.

NeverMarryAStripper
NeverMarryAStripper

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

Posted Images

NeverMarryAStripper

NeverMarryAStripper 21780

Posted
Posted
2 hours ago, hornian said:

Well of course. Their losses are better than our loss. 

I don't think it's necessarily that they are better than ours, they just have twice as many.  One more quality loss and they might crack the top ten.  Meanwhile with our shitty victories we will probably fall out of the top 20.

Link to comment
Share on other sites
satyanash

satyanash 14276

Posted
  • Author
Posted (edited)
36 minutes ago, texifornia said:

we r winnarz

89% post-game win expectancy for Texas... it looks like we overall played a much better game. But Oklahoma State got nine points of turnover luck from the special teams gaffes which contributed to the close margin. Seems to back up the opinions of those who watched.

Also, why is going for it on 4th down and failing not considered a turnover?

Edited by satyanash
Link to comment
Share on other sites
Burt Macklin

Burt Macklin 18988

Posted
Posted
On 9/22/2019 at 9:36 AM, texifornia said:

 

Us being up to 19th after 4 weeks after being 31st in the preseason, especially when our current 19th place ranking is still being heavily weighed down by the preseason ranking shows he was way off on us preseason, not that he'll ever admit it. 

He was also way off on the B12 as a whole, which started the year at an avg of 6.9 and is already up to 12.1, while that 12.1 is again being weighed down by his preseason numbers.  Unsurprisingly, the SEC is already down from 18 to 13.1, while being buoyed by it's preseason rankings. 

36 minutes ago, texifornia said:

we r winnarz

 

A "game decided mainly by who finished drives." Ok, Bill. Based on his own metrics, we had an 89% win expectancy. A more accurate summation would've been Texas largely dominated the game but two muffed punts kept the score much closer than it should have been. Without the muffed punts, we kill OSU in success rate, YPP, and just about all of his other stats in the advanced box score.

On 9/23/2019 at 3:06 AM, texasstrong12 said:

Based on S&P+ we only play one more top 35 offense (OU). 

I think this is about right. The best offenses we face this year are OU, OK State, and LSU. ISU and Baylor might eventually creep into the top 35 but I need to see more. Lots of questions for both. 

Yeah, I think both struggle to finish top 35. ISU just doesn't have the skill talent or explosiveness and I think Baylor's OL will start to get exposed as they face better DLs than SFA, UTSA, and Rice.  Good Lord, Baylor's schedule has been an absolute joke so far.  aggy's even jealous of their OOC scheduling.  

From a UT standpoint, ISU and Baylor don't pose too many problems for us stylistically. Now that Holgo and Kliff are gone, not many teams left do, besides OU of course. I actually think KSU could be a real challenge for us. They have a pretty complex run scheme, and they executed it really well against MSU.  We can't play Dime against them, and it will be a constant challenge for our front 6 to stay disciplined and find their run fits.  I could see KSU actually finishing top 35 with their run game and Skylar Thompson's efficiency.  I'll be really interested to see how they do against OSU. 

 

On 9/23/2019 at 10:05 AM, satyanash said:

Updated win probability chart. SP+ now predicts us to win 7.6 games

spacer.png

Underdogs to all three of ISU, Baylor, and TCU is hilarious.  

Does anyone remember what our preseason win expectancy was?  That number must've been somewhere around 7.  IIRC, Bill's been under on Herman-coached teams by more than 2 wins two of the years and he was under by about .5 wins the other 2, so he's averaging about 1.5 wins under Herman's actual wins per year.  This year has a good chance of increasing Connely's average miss. 

  • Like 1
Link to comment
Share on other sites
texifornia

texifornia 56434

Posted
Posted
7 minutes ago, Burt Macklin said:

Underdogs to all three of ISU, Baylor, and TCU is hilarious.  

Does anyone remember what our preseason win expectancy was?  That number must've been somewhere around 7.  IIRC, Bill's been under on Herman-coached teams by more than 2 wins two of the years and he was under by about .5 wins the other 2, so he's averaging about 1.5 wins under Herman's actual wins per year.  This year has a good chance of increasing Connely's average miss. 

His home/away weighting must be pretty gnarly.

Link to comment
Share on other sites
satyanash

satyanash 14276

Posted
  • Author
Posted
13 minutes ago, Burt Macklin said:

Does anyone remember what our preseason win expectancy was?  That number must've been somewhere around 7.  IIRC, Bill's been under on Herman-coached teams by more than 2 wins two of the years and he was under by about .5 wins the other 2, so he's averaging about 1.5 wins under Herman's actual wins per year.  This year has a good chance of increasing Connely's average miss. 

SP+'s preseason expectancy for us was 6.7 wins.

Link to comment
Share on other sites
Huckleberry

Huckleberry 32685

Posted
Posted (edited)
23 minutes ago, Burt Macklin said:

Us being up to 19th after 4 weeks after being 31st in the preseason, especially when our current 19th place ranking is still being heavily weighed down by the preseason ranking shows he was way off on us preseason, not that he'll ever admit it. 

He was also way off on the B12 as a whole, which started the year at an avg of 6.9 and is already up to 12.1, while that 12.1 is again being weighed down by his preseason numbers.  Unsurprisingly, the SEC is already down from 18 to 13.1, while being buoyed by it's preseason rankings. 

A "game decided mainly by who finished drives." Ok, Bill. Based on his own metrics, we had an 89% win expectancy. A more accurate summation would've been Texas largely dominated the game but two muffed punts kept the score much closer than it should have been. Without the muffed punts, we kill OSU in success rate, YPP, and just about all of his other stats in the advanced box score.

Yeah, I think both struggle to finish top 35. ISU just doesn't have the skill talent or explosiveness and I think Baylor's OL will start to get exposed as they face better DLs than SFA, UTSA, and Rice.  Good Lord, Baylor's schedule has been an absolute joke so far.  aggy's even jealous of their OOC scheduling.  

From a UT standpoint, ISU and Baylor don't pose too many problems for us stylistically. Now that Holgo and Kliff are gone, not many teams left do, besides OU of course. I actually think KSU could be a real challenge for us. They have a pretty complex run scheme, and they executed it really well against MSU.  We can't play Dime against them, and it will be a constant challenge for our front 6 to stay disciplined and find their run fits.  I could see KSU actually finishing top 35 with their run game and Skylar Thompson's efficiency.  I'll be really interested to see how they do against OSU. 

 

Underdogs to all three of ISU, Baylor, and TCU is hilarious.  

Does anyone remember what our preseason win expectancy was?  That number must've been somewhere around 7.  IIRC, Bill's been under on Herman-coached teams by more than 2 wins two of the years and he was under by about .5 wins the other 2, so he's averaging about 1.5 wins under Herman's actual wins per year.  This year has a good chance of increasing Connely's average miss. 

6.9

https://www.espn.com/college-football/story/_/id/27461638/college-football-conference-previews-storylines-using-sp+

It's there in the same article where he declares the SEC East as the second best division in college football partially because Tennessee is a prime second-year leap candidate.

That's #99 Tennessee, if you're curious. Also note the overall ratings system conference rankings at the bottom.

Edited by Huckleberry
  • Like 3
Link to comment
Share on other sites
Burt Macklin

Burt Macklin 18988

Posted
Posted
10 minutes ago, satyanash said:

SP+'s preseason expectancy for us was 6.7 wins.

 

8 minutes ago, Huckleberry said:

6.9

https://www.espn.com/college-football/story/_/id/27461638/college-football-conference-previews-storylines-using-sp+

It's there in the same article where he declares the SEC East as the second best division in college football partially because Tennessee is a prime second-year leap candidate.

That's #99 Tennessee, if you're curious. Also note the overall ratings system conference rankings at the bottom.

Below 7 wins. L O fucking L.  

I don't even think S&P+ is that bad of a metric (though I think his changes this offseason made it worse), but there were a million obvious reasons why Texas was an outlier. Connely's personal leanings make his writing and analysis of his own data terrible.  

Link to comment
Share on other sites
gmr548

gmr548 16463

Posted
Posted
1 hour ago, Burt Macklin said:

Us being up to 19th after 4 weeks after being 31st in the preseason, especially when our current 19th place ranking is still being heavily weighed down by the preseason ranking shows he was way off on us preseason, not that he'll ever admit it. 

He was also way off on the B12 as a whole, which started the year at an avg of 6.9 and is already up to 12.1, while that 12.1 is again being weighed down by his preseason numbers.  Unsurprisingly, the SEC is already down from 18 to 13.1, while being buoyed by it's preseason rankings. 

 

You see a similar trend in ESPN's FPI:

Texas: 24 to 17

Oklahoma State: 37 to 25

Kansas State: 57 to 27; also #8 in team efficiency. (damn shame the OSU-KSU game gets sentenced to ESPN+)

Hell even Kansas: 108 to 97

The one B12 team that has had a somewhat significant drop is WVU, from 58 to 66. Analytics pretty clearly underrated Texas and the B12 conference going into this year and are correcting as the sample size gets larger. Admittedly, I didn't see KSU grading out as good as they are, nor did most.

Link to comment
Share on other sites
Burt Macklin

Burt Macklin 18988

Posted
Posted
31 minutes ago, gmr548 said:

 

You see a similar trend in ESPN's FPI:

Texas: 24 to 17

Oklahoma State: 37 to 25

Kansas State: 57 to 27; also #8 in team efficiency. (damn shame the OSU-KSU game gets sentenced to ESPN+)

Hell even Kansas: 108 to 97

The one B12 team that has had a somewhat significant drop is WVU, from 58 to 66. Analytics pretty clearly underrated Texas and the B12 conference going into this year and are correcting as the sample size gets larger. Admittedly, I didn't see KSU grading out as good as they are, nor did most.

yeah, KSU is the big surprise. I've been extremely impressed with them. Klieman looks like he can really coach. 

Link to comment
Share on other sites
texifornia

texifornia 56434

Posted
Posted
38 minutes ago, HtownHorn said:

KSU is also a veteran squad, with 20/22 upperclassmen starters. 21/22 starters RS'd. Klieman is a good coach, but he wasn't starting from scratch. It's also likely that Snyder's JUCO approach to recruiting rears it's ugly head next year when he loses 13 senior starters.

Excellent. And then it's smooth sailing, it's not like Klieman has any experience maximizing under-recruited Great Plains/Upper Midwest kids.

Link to comment
Share on other sites
JBJ

JBJ 3647

Posted
Posted
On 9/22/2019 at 9:36 AM, texifornia said:

Mine - with model limitations commentary:

1) Ohio State - they haven't played anyone decent, but are fully connected.  Performance against cupcakes weighs less in my model, but there's nothing to weigh less against yet.  That being said, their numbers aren't so great that they are patently unsustainable.

2) Alabama - Pretty much the same as above.

3) Clemson - Fully connected.  A little bit better with A&M in there, but I won't have much to go on past this point.

4) LSU - Not fully connected. They were preseason #4 as well, but their rating was a bit lower (ratings in the top 10 have inflated since preseason) so they should rise maybe to #2 with another week. Alabama and LSU are separated by only 0.5%. I feel most confident where LSU currently sits vs the other top 5.

5) Georgia - Not fully connected.  Arkansas State game inflates their numbers and will continue to get weighed down - waiting for more data.

6) Wisconsin - Not fully connected and they were preseason #46. But the low preseason factor is balanced out by an inflated game against CMU waiting to be down-weighed.  May end up being about right because model limitations are pulling both ways.

7) Auburn - fully connected and highest-confidence of the top 10.

8: Oregon - not fully connected and lowest confidence of the top 10.  They will fall from here and that might actually also mess with Auburn a bit.

9) Texas - fully connected, high confidence.  Best team in the nation, imo.

10) Oklahoma - very low connectivity (SDU plus a bye).  But playing really close to their preaseason rating anyway.  Could go up or down because data is very limited and unreliable here.

 

Note: "high confidence" is only really relative to other teams at this point in the season.  The model is still overall 'young.'  2 weeks from now pretty much every team will be better settled than the most settled team at the moment.

Link to comment
Share on other sites
Treefidy

Treefidy 1463

Posted
Posted
On 9/24/2019 at 12:39 PM, Huckleberry said:

6.9

https://www.espn.com/college-football/story/_/id/27461638/college-football-conference-previews-storylines-using-sp+

It's there in the same article where he declares the SEC East as the second best division in college football partially because Tennessee is a prime second-year leap candidate.

That's #99 Tennessee, if you're curious. Also note the overall ratings system conference rankings at the bottom.

So the SEC has 14 of the hardest 16 SOS, 1-13 and 16.   Right.  

What a complete fucking tool.  Exhibit 2,724 of why people hate the SEC and the media bias.   Exhibit 2,725 is aggy ranked at 2-2 with wins over power programs TxSt and Lamar.  

  • Like 1
Link to comment
Share on other sites
gmr548

gmr548 16463

Posted
Posted
So the SEC has 14 of the hardest 16 SOS, 1-13 and 16.   Right.  
What a complete fucking tool.  Exhibit 2,724 of why people hate the SEC and the media bias.   Exhibit 2,725 is aggy ranked at 2-2 with wins over power programs TxSt and Lamar.  

You think aggy's wins are good? Wait til you hear about their losses!

Sent from my SM-G920V using Tapatalk

Link to comment
Share on other sites
satyanash

satyanash 14276

Posted
  • Author
Posted
Quote

Strong passing grades

251.png?w=110&h=110&transparent=true

Texas (3-1)
Current FPI title odds: <0.1% (preseason: same)

Preseason Ifs ...
If ... Texas can figure out how to run the ball without getting Sam Ehlinger hit so much
If ... big-play blue-chippers can actually make big plays
If ... a super-young secondary is ready to not only hold the fort, but improve
If ... a Tom Herman team can play every game as an underdog

Early developments have been mostly positive for a Texas team that voters liked a lot more than analytics. Longhorns running backs are generating efficiency, and they're benefitting even further from the emergence of a ridiculously efficient receiver in Devin Duvernay (87% catch rate, 67% success rate). UT is also getting a massive big-play boost from sophomore wideout Brennan Eagles (10 catches, 276 yards, four TDs). This has created a far greater level of offensive consistency, and it has helped Texas to play well as a favorite as well as an underdog.

The bad news: That young secondary is getting obliterated by injury. Safeties Caden Sterns and B.J. Foster are both out indefinitely, as are three other defensive backs. The Longhorns are on a bye this week, thankfully, and take on a less-than-amazing West Virginia passing game in Week 6, but Oklahoma looms on the schedule.

 

Link to comment
Share on other sites
sushihorn

sushihorn 3233

Posted
Posted
Quote

Early developments have been mostly positive for a Texas team that voters liked a lot more than analytics. Longhorns running backs are generating efficiency, and they're benefitting even further from the emergence of a ridiculously efficient receiver in Devin Duvernay (87% catch rate, 67% success rate). UT is also getting a massive big-play boost from sophomore wideout Brennan Eagles (10 catches, 276 yards, four TDs). This has created a far greater level of offensive consistency, and it has helped Texas to play well as a favorite as well as an underdog.

Like most Fantasy Football guys, Connelly over-focuses on the eligible players on offense and almost completely ignores OL play.  I suspect that's because there aren't many stats to measure the success of the line except indirectly.  So it's a stat geek version of the drunk looking for his keys under the same streetlamp because that's where the light is.

I'm sure he's undervaluing the Offensive Line.  If the logical conclusions from that hold, it's pretty clear why he can't correctly value Tom Herman teams.  He's looking at skill position production as a partial proxy for something he can't measure (OL effectiveness).  But because he can't (or at least won't) separate line yards from RB yards he simplistically attributes 100% of the yardage to the back.  Thus the loss of a plurality of Texas' running game "production" with the departure of Tre Watson from 2018.  That type of assessment is always going to undervalue a rising program and especially one driven by large improvements in the play of the O-Line.

His system completely misses the fact that Texas returned its two best starters (Cosmi and Shack).  It doesn't account for adding a GT all-conference Guard.  And it cannot see (more excusably) the value of a RS freshman breaking out at the other Guard position - just like it missed Cosmi doing the same at Tackle last season.  Connelly's code sees "2 returning starters" not 4 veterans (all 4 good and 2 great) and one young beast.

The systemic flaw here is Billy's inability to measure something critical and the refusal to admit that it is is critical.  That arrogance is compounded by incorrectly attributing most of that critical factor to a result, not a cause (RB yardage).  It is further aggravated by assuming that the OL blocking which constitutes so much of that production leaves with the runner.  To top it all off, he uses only the crudest measure of OL quality which leads to wildly inaccurate conclusions with coaching staffs that have a history of producing high-quality OL regularly and especially when they are starting from a low base from which major improvement is likely each year for some time.  Connelly's model works fine for steady state programs where the future looks pretty much like the past.  Where there is a significant change from year to year his model will always be behind the informed humans that actually watch the game.

  • Like 2
Link to comment
Share on other sites
Walden Ponderer

Walden Ponderer 12639

Posted
Posted
22 minutes ago, sushihorn said:

He's looking at skill position production as a partial proxy for something he can't measure (OL effectiveness).  But because he can't (or at least won't) separate line yards from RB yards he simplistically attributes 100% of the yardage to the back.  Thus the loss of a plurality of Texas' running game "production" with the departure of Tre Watson from 2018.  That type of assessment is always going to undervalue a rising program and especially one driven by large improvements in the play of the O-Line.

This is far and away the biggest flaw in most analyses of college football, not just Connelly's.

The OL and DL are the most important positions on a team, and for the most part, not in ways that are directly measurable. Therefore, the best way to gauge their effectiveness is to use statistics traditionally credited just to individuals. But it's a team game, first and foremost.

The irony is, that makes part of his model more sensible than it at first glance appears. To truly gauge the strength or weakness of a team, you really do need to know what their long-term growth is like, and that means collecting data from the last 3, 4, or 5 years. The problem is, you have to know what you are looking for from those time frames. If you're looking for returning experience on the lines, and maybe tracking declining body fat %, then you might be on to something. If you're just tracking returning starts and recruiting rankings, though, then probably not.

Link to comment
Share on other sites
Burt Macklin

Burt Macklin 18988

Posted
Posted (edited)
1 hour ago, sushihorn said:

Like most Fantasy Football guys, Connelly over-focuses on the eligible players on offense and almost completely ignores OL play.  I suspect that's because there aren't many stats to measure the success of the line except indirectly.  So it's a stat geek version of the drunk looking for his keys under the same streetlamp because that's where the light is.

I'm sure he's undervaluing the Offensive Line.  If the logical conclusions from that hold, it's pretty clear why he can't correctly value Tom Herman teams.  He's looking at skill position production as a partial proxy for something he can't measure (OL effectiveness).  But because he can't (or at least won't) separate line yards from RB yards he simplistically attributes 100% of the yardage to the back.  Thus the loss of a plurality of Texas' running game "production" with the departure of Tre Watson from 2018.  That type of assessment is always going to undervalue a rising program and especially one driven by large improvements in the play of the O-Line.

His system completely misses the fact that Texas returned its two best starters (Cosmi and Shack).  It doesn't account for adding a GT all-conference Guard.  And it cannot see (more excusably) the value of a RS freshman breaking out at the other Guard position - just like it missed Cosmi doing the same at Tackle last season.  Connelly's code sees "2 returning starters" not 4 veterans (all 4 good and 2 great) and one young beast.

The systemic flaw here is Billy's inability to measure something critical and the refusal to admit that it is is critical.  That arrogance is compounded by incorrectly attributing most of that critical factor to a result, not a cause (RB yardage).  It is further aggravated by assuming that the OL blocking which constitutes so much of that production leaves with the runner.  To top it all off, he uses only the crudest measure of OL quality which leads to wildly inaccurate conclusions with coaching staffs that have a history of producing high-quality OL regularly and especially when they are starting from a low base from which major improvement is likely each year for some time.  Connelly's model works fine for steady state programs where the future looks pretty much like the past.  Where there is a significant change from year to year his model will always be behind the informed humans that actually watch the game.

Long post alert. Tl;dr: His systems’ not bad, but his individual analysis based on the system’s outputs is just plain horrible. This is exacerbated when applied to UT, because UT’s roster from this year to last year and Herman/TO’s coaching philosophies make UT a massive outlier in Connely’s system, but he’s too arrogant and/or obstinate to recognize it.

The flaw in Connely’s analysis of individual teams is that he created a system that can be applied to every team, to be used for betting/ranking even if you’ve never watched that team. As such, metrics like returning production on defense and returning linemen on O are reasonably solid metrics to use, but there will obviously be exceptions, like poor defensive starters leaving or adding a GT and/or young OL who are much more talented than the ones they replace.  Overall, his system isn’t bad at betting lines/predicting team performance from a 30,000 foot view. There are some weight issues and metric he uses I don’t like, but the system has performed decently well compared to other public analytics for the same use.

The real issue with Connely is he acts like his system (and his related analysis) allows him to have a great knowledge of basically every single team.  Texas’ roster this year and Herman/Orlando’s coaching styles are both massive outliers in his system. Rather than recognize that and admit S&P will sometimes struggle with these kind of outliers, he just constantly doubles down as if he knows what he’s talking about. No roster had as great of a disparity in talent level of departing players compared to talent level of incoming players as UT from 2018 to 2019. Guys like Vahe, Hager, Wheeler weren’t good and were poor system fits. Their replacements have all been massive steps up for the team while S&P just counted all three of those as production lost. Similarly, no team has had as big of a change in quality of coaching as Strong-->Herman, so looking at the last 5 years of UT’s performance will also be extremely misleading.

Finally, from a coaching perspective, both Herman and Orlando’s philosophies don’t match up with what Connely has built his system to value: maximizing production on a per play basis. It’s not necessarily a bad way to build the system, but it does miss some things. For instance, while our offense is obviously more explosive this year, Herman still believes in using the offense to control the game and the clock while simultaneously protecting his defense. Herman’s not trying to score in 3 plays every drive. Furthermore, Herman’s use of power running and QB running means he can convert things like 3rd and short at a higher clip than most offenses, but a system like S&P+ primarily just discounts this higher a conversion rate as a statistical outlier that will regress to the mean instead of potentially being a dynamic of Herman’s system that could consistently outperform the average on a yearly basis. Same thing with Orlando. TO wants to create drive-killing plays to limit drive efficiency instead of limiting offensive production on a per play basis, and he’s willing to accept that some explosive plays and 5 or less play TD drives will occur as a result. I’m not saying TO’s philosophy is completely right, and the LSU game certainly showed the primary flaws with his phiosiphy, but it’s clear TO defenses don’t value the same thing as S&P+, which leads to S&P+ traditionally undervaluing TO’s defenses compared to other metrics.

For instance, FEI (one of the main competitors to S&P+) values possession efficiency over per play efficiency, whereas S&P+ puts a higher value on per-play efficiency. I put this in another thread, but FEI has had Orlando’s defenses ranked higher than S&P+ has 9 out of 11 years and many times it’s been a pretty big difference.  Overall, FEI has had Herman/Orlando teams ranked higher than S&P+ every single year by a wide margin (2018: 30 v. 16, 2017: 50 v. 35, 2016: 39 v. 26, 2015: 44 v. 10). That’s an average of 21 spots difference in the two systems’ rankings per year. Obviously, that transcends a statistical anomaly year or specific roster construction outliers. Fundamentally, Herman/Orlando value different factors than S&P+, which leads to S&P+ underrating their teams every single year. Rather than recognize this and just admit it’s an established exception, he doubles down every single year that his system is right and Herman/TO value the wrong things/aren’t as good as the eye test and other analytics say they are.

24 minutes ago, Walden Ponderer said:

This is far and away the biggest flaw in most analyses of college football, not just Connelly's.

The OL and DL are the most important positions on a team, and for the most part, not in ways that are directly measurable. Therefore, the best way to gauge their effectiveness is to use statistics traditionally credited just to individuals. But it's a team game, first and foremost.

The irony is, that makes part of his model more sensible than it at first glance appears. To truly gauge the strength or weakness of a team, you really do need to know what their long-term growth is like, and that means collecting data from the last 3, 4, or 5 years. The problem is, you have to know what you are looking for from those time frames. If you're looking for returning experience on the lines, and maybe tracking declining body fat %, then you might be on to something. If you're just tracking returning starts and recruiting rankings, though, then probably not.

I see two potential answers to this.  Rely on a service like PFF and use their player grades instead of returning production on defense and returning starts on the OL. But this requires a lot of faith in PFF, whose grades seems much less reliable at the college level compared to the NFL.

The other option is to look at past performance. He tries to do this through looking at a team’s last 5 years’ performance, but that has obvious issues when the current HC wasn’t there for 5 years (like weighing down this year’s teams because of how bad Charlie was 2014-2016). It would be more interesting if he isolated each coach’s last 5 years performance (he’d have to use his own system which would make it somewhat circular but probably still better than what he’s using now), whether as a HC or coordinator, and used that to give each coach a numerical rating, which he applied to the team’s ranking, instead of using the last 5 years of rankings for the school.

Edited by Burt Macklin
  • Like 4
Link to comment
Share on other sites
Walden Ponderer

Walden Ponderer 12639

Posted
Posted
31 minutes ago, Burt Macklin said:

I see two potential answers to this.  Rely on a service like PFF and use their player grades instead of returning production on defense and returning starts on the OL. But this requires a lot of faith in PFF, whose grades seems much less reliable at the college level compared to the NFL.

The other option is to look at past performance. He tries to do this through looking at a team’s last 5 years’ performance, but that has obvious issues when the current HC wasn’t there for 5 years (like weighing down this year’s teams because of how bad Charlie was 2014-2016). It would be more interesting if he isolated each coach’s last 5 years performance (he’d have to use his own system which would make it somewhat circular but probably still better than what he’s using now), whether as a HC or coordinator, and used that to give each coach a numerical rating, which he applied to the team’s ranking, instead of using the last 5 years of rankings for the school.

I actually like that idea, even with the circularity, simply because it replaces a currently recursive non-sequitur with a recursive sequitur, so to speak. Instead of comparing apples to monkey wrenches, like his current system does, it would be comparing fried apple pies to apples. The data on linemen would at least be in the same universe as the other data, even if it still wasn't technically on point.

Link to comment
Share on other sites
Burt Macklin

Burt Macklin 18988

Posted
Posted
2 hours ago, texifornia said:

Updated win expectancy chart - the big movement is that we're now predicted to beat Baylor.

 

We should still be favored against ISU and TCU, but it is funny that we become favorites against Baylor on a week where we don’t play but they beat ISU, while we’re still underdogs against ISU in his system. 
 

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.

×
 Share
Go to topic listing


Football ... Basketball ... Baseball ... Other Sports ... Futbol ... 🤫995🤫 ... Gambling ... Movies & TV ... Music ... Hobbies ... Lulz ... Food & Travel ... Daily Texan ... Business and Markets ... Cloak Room ... Help ... For Sale ... Board Discussion ... Subscribe!... Donate!... Advertise... COOKIE MONSTER!

×
×
  • Create New...