2
Started By
Message
re: Week 8 Power Ratings and SEC slate
Posted on 10/22/26 at 8:47 am to SparkyWilson
Posted on 10/22/26 at 8:47 am to SparkyWilson
quote:
Beating bad teams by covering the spread or better doesn't make a team a good team.
Correct. Blowing out a bad team doesn't credit you that much. LSU blowing out LA Tech gives them a fraction of the credit of LSU blowing out TAMU.
The problem is you are assuming Clemson and TAMU are "bad" teams. They are not.
quote:
Somewhere in all this is the assertion that LSU is a good team.
You can say "good", the model doesn't know what "good" means. It's about relative performance to the rest of the field. LSU has 3 potent (blowout) wins against top 50 teams and has two close losses. This is also why Bama is high - they have 3 potent wins against top 50 teams and two losses. One of those losses was close at home, the other was not so close - so they are a little bit lower than LSU.
Posted on 10/22/26 at 9:57 am to MilkJug
quote:
Correct. Blowing out a bad team doesn't credit you that much. LSU blowing out LA Tech gives them a fraction of the credit of LSU blowing out TAMU.
The problem is you are assuming Clemson and TAMU are "bad" teams. They are not.
They are only decent teams based how your model considers them. If the model was adjusted for the P4 only (I understand why you don't), they wouldn't be "good" teams.
quote:
You can say "good", the model doesn't know what "good" means. It's about relative performance to the rest of the field. LSU has 3 potent (blowout) wins against top 50 teams and has two close losses. This is also why Bama is high - they have 3 potent wins against top 50 teams and two losses. One of those losses was close at home, the other was not so close - so they are a little bit lower than LSU.
You just said that TAMU and Clemson aren't "bad" teams, which I take to mean you model doesn't consider them bad teams. So it "knows" when a team is bad but not good?
And the potent wins thing, that again propped up by the disparity in talent when considering the entire FBS. And the model doesn't consider other factors like not having KamarioTaylor for LSU (I don't care about the other injuries on both sides; the QB is the single most important position). Your model only sees that they won by 30 points and it still considers Mississippi State a "good" team because of their body of work at the time of the game but that body of work was heavily a result of Kamario Taylor.
And does the model adjust the ranking based on how teams do in the future. E.g. if TAMU struggles down the stretch does that potent win by LSU as you call it get devalued because they weren't as "good" as the model indicated at the time? or is that locked in the week it happens? Same if Kamario Taylor comes back and if (and I emphasize if) Mississippi State starts winning again, does that give LSU's win more weight?
Anyway, no need to go round and round if you want to; I don't expect you to spend more time on this unless you just want to. I get the gist of what it's purpose is, it's just not something I find useful for me.
Posted on 10/22/26 at 11:43 am to SparkyWilson
quote:
They are only decent teams based how your model considers them. If the model was adjusted for the P4 only (I understand why you don't), they wouldn't be "good" teams.
Even if I removed all group of 6 teams, the relative difference between the Power 4 teams wouldn't change.
The model measures Clemson and TAMU's performance the same way it measures every other Power 4 team.
quote:
You just said that TAMU and Clemson aren't "bad" teams, which I take to mean you model doesn't consider them bad teams. So it "knows" when a team is bad but not good?
No, it doesn't think in terms of "good" and "bad". I said they're not a "bad" team in response to you.
quote:
Your model only sees that they won by 30 points and it still considers Mississippi State a "good" team because of their body of work at the time of the game but that body of work was heavily a result of Kamario Taylor.
Stop saying "good" or "bad" - it means nothing. I admitted earlier that injuries do effect the result of a single game, but over the course of a season, factoring in injuries to a score prediction model hurts more than it helps. The reasons for this is what I've already explained - because:
1. there's no way to measure the delta between the starting QB and the backup QB.
2. the other team has injuries as well - IT MATTERS
Examples of #1:
Florida's QB Philo goes out and their backup comes in - the backup does much better.
Miss St's Taylor goes out and their backup comes in - the backup is much worse.
If the model docked a team for beating a team without their starting QB, then LSU would be docked and Texas would be docked even more for losing to a backup QB. Do you see why this is bad?
quote:
And does the model adjust the ranking based on how teams do in the future. E.g. if TAMU struggles down the stretch does that potent win by LSU as you call it get devalued because they weren't as "good" as the model indicated at the time?
Absolutely. As I said in my previous long post, every game is re-evaluated every week. If TAMU start playing shittier, then LSU will be hurt by that.
quote:
I get the gist of what it's purpose is, it's just not something I find useful for me.
This is how models work... My model produces similar results to Vegas - because Vegas uses a similar model.
This post was edited on 10/22/26 at 11:48 am
Posted on 10/22/26 at 12:51 pm to MilkJug
Had a bunch of stuff written; really wouldn't be helpful, so I decided to pare it down to this:
1. LSU has a potent win vs MSU by virtue of:
A. Beating the spread by 16.5 points
B. MSU body of work to that point (they covered in wins vs Minn(72),
SoCar(29), and Mizzou(27) (ESPN FPI in parenthesis since you
said your model gets close).
C. Offense and defense ranks
2. MSU body of work was largely due to Kamario playing.
3. Kamario doesn't play against LSU
Given those:
Do the future predicted spreads for LSU (or any team with similar circumstances) continue to be affected by a game like this? Like if Kamario comes back and State starts winning, it will give a boost to LSU's predicted spread. As in, the model considers that win more potent as the season goes on if State wins more games. Seem like that would be an issue for overpredicting a teams spread.
1. LSU has a potent win vs MSU by virtue of:
A. Beating the spread by 16.5 points
B. MSU body of work to that point (they covered in wins vs Minn(72),
SoCar(29), and Mizzou(27) (ESPN FPI in parenthesis since you
said your model gets close).
C. Offense and defense ranks
2. MSU body of work was largely due to Kamario playing.
3. Kamario doesn't play against LSU
Given those:
Do the future predicted spreads for LSU (or any team with similar circumstances) continue to be affected by a game like this? Like if Kamario comes back and State starts winning, it will give a boost to LSU's predicted spread. As in, the model considers that win more potent as the season goes on if State wins more games. Seem like that would be an issue for overpredicting a teams spread.
Posted on 10/22/26 at 1:00 pm to HYDRebs
quote:
Anyone with a brain knows that was an arse whooping
It was tied 24-24 in the 4th qtr and you won by 8. You need to learn what an arse whooping is
Posted on 10/22/26 at 1:57 pm to skrayper
Analytics are dogshat.
Posted on 10/22/26 at 4:31 pm to SparkyWilson
quote:
Do the future predicted spreads for LSU (or any team with similar circumstances) continue to be affected by a game like this?
They do. But if LSU is overrated this week then they wont live up to their rating and they will fall. Does that make sense? And this goes for every team. It works itself out.
Again, over the course of a season, injuries don't make a big difference when pairing two teams together - they are insignificant. Because all teams are dealing with injuries.
Just for you, I tried docking LSU four different ways for Taylor being out and they all made the prediction model's MAE worse.
Also, the credit for any one game falls off by 5% per week but does not go lower than 70%. The model cares a little bit more about "what have you done lately". So by the end of the season, that MSU game won't carry the same weight that it does this week. The Clemson and TAMU games also wont carry as much weight. The Clemson game currently carries 70% weight. The reason the model works this way is (1) teams improve throughout the season and (2) it improves MAE.
This post was edited on 10/22/26 at 4:35 pm
Posted on 10/22/26 at 5:25 pm to MilkJug
quote:
Correct. Blowing out a bad team doesn't credit you that much.
Really?
Because that is exactly what your "Pretender Rankings" show. Look at the top 3 teams in your rankings.
The problem is that when you play a team that is actually good, they don't have that same "Power."
Prove me wrong.
Posted on 10/22/26 at 5:28 pm to MilkJug
It's about wins and losses. What the frick is this post? This means nothing.
Posted on 10/22/26 at 5:32 pm to MilkJug
quote:
They do. But if LSU is overrated this week then they wont live up to their rating and they will fall. Does that make sense? And this goes for every team. It works itself out.
Yes. For clarification, if Mississippi State started winning, then that LSU win continues to be a boost to LSU in this model regardless of how they perform in the weeks after. Is that correct?
quote:
Again, over the course of a season, injuries don't make a big difference when pairing two teams together - they are insignificant. Because all teams are dealing with injuries.
Can you expound on that? Maybe you stated this already, but are spreads not influenced by injury at all? I would think a suddenly healthy Kamario would shift the spread for LSU v MSU. Or if Sam Leavitt suddenly was unavailable against Auburn, I would think that spread would swing towards Auburn.
Now, that would resolve itself by kickoff, and I don't know what a model could do to say that player(s) X being out equals this many points. Whatever that is set to, then it's a matter of the final score versus the predicted spread.
quote:
Just for you, I tried docking LSU four different ways for Taylor being out and they all made the prediction model's MAE worse.
Thanks but not necessary. I'm not concerned anymore with if it should have been reflected in the rating as of this week based on this discussion. If the spread already accounts for injury, then it's a matter of what amount was accounted for in the spread based on the perceived value of the injury and then how the team performs against the predicted spread.
it sounds like what you're saying is if Team A has a potent win against Team B, and Team B goes on to win more games, Team A's rating will get to benefit from that even if there was a significant variable present for Team B winning more games that wasn't present when they lost to Team A.
Posted on 10/22/26 at 6:19 pm to SparkyWilson
quote:
For clarification, if Mississippi State started winning, then that LSU win continues to be a boost to LSU in this model regardless of how they perform in the weeks after. Is that correct?
Yes, but again - the Miss St wins weighs less as the rest of the season progresses - also, one game doesn't define a team's rating - its based on the entire season. LSU's +16 or whatever in that game is a small piece of their rating. You are putting too much weight on this one game.
quote:
Can you expound on that? Maybe you stated this already, but are spreads not influenced by injury at all? I would think a suddenly healthy Kamario would shift the spread for LSU v MSU. Or if Sam Leavitt suddenly was unavailable against Auburn, I would think that spread would swing towards Auburn.
I didn't say they don't effect the spread - I said the opposite, they do effect a single game but over the course of a season, they are insignificant. And they don't even effect a single game that much. The Miss St game opened at 10.5 and climbed to 17.5 - that's a touchdown difference. In the grand scheme of things, that's not a big difference. And LSU scored almost twice that margin. My model's spread on the game was 12
quote:
it sounds like what you're saying is if Team A has a potent win against Team B, and Team B goes on to win more games, Team A's rating will get to benefit from that even if there was a significant variable present for Team B winning more games that wasn't present when they lost to Team A.
Team A's rating does benefit but it's not as much as you may think. Not only does it end up being weighted by 70%, it is also in a pool of 12-16 other games for Team A. Even If I cut LSU's credit in half for the Miss St game, they may only fall 2 points in the ratings - if that.
Posted on 10/22/26 at 6:27 pm to MilkJug
quote:
Team A's rating does benefit but it's not as much as you may think. Not only does it end up being weighted by 70%, it is also in a pool of 12-16 other games for Team A. Even If I cut LSU's credit in half for the Miss St game, they may only fall 2 points in the ratings - if that.
I'm not worried about the immediate impact, just does it continue to be a benefit to whatever degree as season progresses. Like if a team all of sudden starts tanking how long does it take for those potent wins to stop impacting future spreads.
Posted on 10/22/26 at 8:07 pm to SparkyWilson
quote:
I'm not worried about the immediate impact, just does it continue to be a benefit to whatever degree as season progresses. Like if a team all of sudden starts tanking how long does it take for those potent wins to stop impacting future spreads.
Miss St is currently a 1.5 point favorite over Oklahoma at home this weekend. If they go out and lose by 2 touchdowns, it will immediately effect LSU's rating. Every game played matter every week.
Posted on 10/22/26 at 9:19 pm to PurpleandGeauld
quote:
LSU played without Trey'Dez Green, Whit Weeks, and 2 other starters. Princewill was out over half the game as well iirc.
I know this may or may not have been enough to change things BUT I was closely watching Bama and their injuries to the Center and RCW. The re-shuffled O-Line seemed off the whole game for us plus there is no doubt not having RCW gave UT’s D some abilities to take more liberties in their defensive game plan…but yes injuries can take a toll but at least your team was able to get by in spite of them.
Posted on 10/23/26 at 7:38 am to Crimson1st
quote:
I know this may or may not have been enough to change things BUT I was closely watching Bama and their injuries to the Center and RCW. The re-shuffled O-Line seemed off the whole game for us plus there is no doubt not having RCW gave UT’s D some abilities to take more liberties in their defensive game plan…but yes injuries can take a toll but at least your team was able to get by in spite of them.
Yes, injuries matter, but every team deals with them.
LSU not having DeShawn Spears on defense completely changes what they can do on defense schematically. Then you take Weeks out, he's the QB on defense. That's a massive hit. And it showed itself in the Kentucky game.
LSU not having Green on offense completely changes what they can do on offense too.
But every team has injuries. You can't just dock one team in a game because the other team had an injury - I know that may sound counter-intuitive, but it's true.
Look at some of the SEC injury reports for this week - it's insane.
Loading Twitter/X Embed...
If tweet fails to load, click here. The main reason why it's not realistic is this:
(Trey'dez Green - Roman Mothershead) = ??
(Deshawn Spears - Ty Benefield at nickel) = ??
(Ty Benefield - Treylan James at safety) = ??
(Whit Weeks - TJ Dottery at mike) = ??
(TJ Dottery - Davhon Keys at OLB) = ??
then Miss St
(Kamario Taylor - Puckett) = ??
and their other injuries..
there's no way you could even come close to figuring that out. Because we have no data on Mothershead, James, Keys, and Puckett. It would just be human guessing - which is the worst possible thing you can do for a score prediction model.
This post was edited on 10/23/26 at 7:40 am
Posted on 10/23/26 at 1:45 pm to MilkJug
quote:
Miss St is currently a 1.5 point favorite over Oklahoma at home this weekend. If they go out and lose by 2 touchdowns, it will immediately effect LSU's rating. Every game played matter every week.
Got you.
I've certainly beaten the horse to death, and you've been more than accommodating and patient. I sincerely appreciate it. Cheers.
Page 8 of 8
Popular
Back to top
Follow SECRant for SEC Football News