Stats models and rankings

iceeater1969

Helluva Engineer
Messages
8,774
Finally got a decent OL coach.
And we have an oc + qb that get the ball out to receiver in flat. Both have complimented each other AND WHEN WE RUN ( wo rpo) on slant we get good yards.

Last game they did stuff to counteract the oc qb flat passes= sent dl full force ,had db up on our slower wr, had lb or a de on qb, if qb had time the db was all over wr unless he made a move ( wr Singleton and King out of sync, or wr made no move and king appesred to throw to thier Safety

We , hopefully have adjudted tova more complicated ( te or wr over middle for example, qb looks one way then shifts and throws the other way) passing attack, and give the ball to Dontae on full go quick hitter.
 

slugboy

Moderator
Staff member
Messages
10,100
Hooo boy, I’m starting something with this one:


IMG_2426.jpeg


Collins didn’t inherit that bad of a talent situation. He increased the recruiting scores but tanked the record.
Key had (theoretically) talent to work with. The chart doesn’t show how hard he got hit by the portal (he did well there, too—King, for example)
I accept the irony that I’m posting a chart with the talent composite as one axis, since that’s a “stat” I complain about.
Hats off to Elko and Leopold. Poor Babers—he started in a bad spot and had a big turnaround job
 

stinger78

Ramblin' Wreck
Messages
736
Great chart. I also contend with the "talent" metric as talent cannot be discussed apart from system, as talent and system go hand-in-hand. IOW, you recruit to your system. That said, there were some deficits relative to what TFG wanted to do. The mark of a good HC when he comes in is how well he is able to fit the talent to his new system. CPJ did that masterfully, IMPO, while TFG made a total hash of it. To me, therein is a major difference between the two - CPJ made it work (20 wins his first two seasons) while TFG made it an excuse (6 wins his first two seasons).
 

yeti92

Helluva Engineer
Messages
2,336
Team talent composite is not a very useful metric for measuring anything even remotely detailed. If players were rerated every year it might be a little better, but as it stands, everything is based on their high school ratings which we all know are regularly wrong, sometimes by a lot. Guys like Tariq Carpenter and Keion White, despite being NFL-level talent, count as 2 stars towards the composite, and guys like Antonneous Clayton count as 4/5 stars despite never coming anywhere close to that level of performance.
 

leatherneckjacket

Helluva Engineer
Messages
1,424
Location
Atlanta, GA
Team talent composite is not a very useful metric for measuring anything even remotely detailed. If players were rerated every year it might be a little better, but as it stands, everything is based on their high school ratings which we all know are regularly wrong, sometimes by a lot. Guys like Tariq Carpenter and Keion White, despite being NFL-level talent, count as 2 stars towards the composite, and guys like Antonneous Clayton count as 4/5 stars despite never coming anywhere close to that level of performance.
At an individual level it is highly inaccurate, but at an aggregate level it is fairly accurate and directional.
 

ibeattetris

Helluva Engineer
Messages
3,488
The fact that everyone talks so favorable of Elmo just goes to show how irrelevant strength of schedule is. Duke plays and beats nobody and they are crowned as something beyond what they are. The only team they beat was a down bad Clemson who was truly atrocious to start the year. I’ve never seen a team more lauded for accomplishing little. They pretty much did the same thing last year, winning a bunch of game against nobody’s. I wonder what people would have said about Collins if he’d have gotten to skate by on easy Duke schedules.
 

ibeattetris

Helluva Engineer
Messages
3,488
At an individual level it is highly inaccurate, but at an aggregate level it is fairly accurate and directional.
Problem though is our talent composite was shifted towards skill guys on offense and defense. We did not have the right players to run the offense Collins forced upon us making it seem so much worse.
 

JacketOff

Helluva Engineer
Messages
2,706
The fact that everyone talks so favorable of Elmo just goes to show how irrelevant strength of schedule is. Duke plays and beats nobody and they are crowned as something beyond what they are. The only team they beat was a down bad Clemson who was truly atrocious to start the year. I’ve never seen a team more lauded for accomplishing little. They pretty much did the same thing last year, winning a bunch of game against nobody’s. I wonder what people would have said about Collins if he’d have gotten to skate by on easy Duke schedules.
Because Duke has been playing the same weak schedules for decades and not winning at the same rate that Elko has been able to. He took over a team that went 3-9 and immediately went 9-3 against a similar schedule. It was their first winning record since 2018, and their first winning record in the ACC since 2015. They have a pretty good chance to have back to back .500 or greater ACC records for the first time since Cutcliffe did it 3 times in a row from 2013-15. Before that stretch, their last winning record in the ACC was 1994.

Their schedule this year is much tougher than last year’s hence the drop in wins. They played 3 of the top 4 ACC teams (Tech is the only one they missed), plus Clemson, plus Notre Dame, plus UNC. As a college coach you get judged based on what your predecessors did. They are the ones that set the standard for your expectations. Cutcliffe had some solid years based on Duke’s history, especially after taking over for Ted Roof who went 4-42 in 4 full seasons. But I would say nobody has made such a big impact so quickly at Duke as Elko has since Steve Spurrier did in the 80s.

Collins didn’t get the benefit of Duke’s scheduling, but he did inherit a much better overall program with much higher expectations than what he would have at Duke. The historical standard for Tech coaches is to win 7 games a year. If you can do that, you’ll be considered “alright.” Fail to do so and you’ll be considered a failure, do more than that and you’ll be considered a great. Not all schedules are created equally, but neither are all programs or their expectations.
 

stinger78

Ramblin' Wreck
Messages
736
Because Duke has been playing the same weak schedules for decades and not winning at the same rate that Elko has been able to. He took over a team that went 3-9 and immediately went 9-3 against a similar schedule. It was their first winning record since 2018, and their first winning record in the ACC since 2015. They have a pretty good chance to have back to back .500 or greater ACC records for the first time since Cutcliffe did it 3 times in a row from 2013-15. Before that stretch, their last winning record in the ACC was 1994.

Their schedule this year is much tougher than last year’s hence the drop in wins. They played 3 of the top 4 ACC teams (Tech is the only one they missed), plus Clemson, plus Notre Dame, plus UNC. As a college coach you get judged based on what your predecessors did. They are the ones that set the standard for your expectations. Cutcliffe had some solid years based on Duke’s history, especially after taking over for Ted Roof who went 4-42 in 4 full seasons. But I would say nobody has made such a big impact so quickly at Duke as Elko has since Steve Spurrier did in the 80s.

Collins didn’t get the benefit of Duke’s scheduling, but he did inherit a much better overall program with much higher expectations than what he would have at Duke. The historical standard for Tech coaches is to win 7 games a year. If you can do that, you’ll be considered “alright.” Fail to do so and you’ll be considered a failure, do more than that and you’ll be considered a great. Not all schedules are created equally, but neither are all programs or their expectations.
I certainly haven't analyzed their talent, but I think Cutcliffe left Elko some talent when he retired. Much more so than Roof left him.
 

slugboy

Moderator
Staff member
Messages
10,100
I certainly haven't analyzed their talent, but I think Cutcliffe left Elko some talent when he retired. Much more so than Roof left him.
Tech is a hard job. There’s another thread on Jamey Chadwell, and he took a much easier job at Liberty. The theme I see there is that we didn’t put together enough money to hire him, but I think he probably wanted the much easier job and would have taken a lot less for that.*

Elko had a hard job at Duke, and he’s done well. Cutcliffe may have left him some pieces, but the team didn’t seem to be in great shape to me. I’d say he’s been incredibly successful. That doesn’t mean he’d have succeeded here, in a tougher environment. And even though we’ve had more success than Duke overall, I do think they’re easier in a lot of a ways now—enough support, not as much pressure, easier schedule, more flexible majors, etc.

I was going to pull some analytics, but thought people might get more out of this:


*I didn’t think he’d fit in at Liberty, and he’s already looking at Mississippi State.
 

roadkill

Ramblin' Wreck
Messages
839
Now, I need to clean this up with some stats. Maybe this is why the “make the schedule easier” vs “you gotta schedule big teams to build a program” argument never settles out.
So, from these charts, should we conclude that if the data is properly parsed along the lines of how teams actually schedule games (G5-P5), our resume is more likely to get worse with a harder schedule?
 

slugboy

Moderator
Staff member
Messages
10,100
So, from these charts, should we conclude that if the data is properly parsed along the lines of how teams actually schedule games (G5-P5), our resume is more likely to get worse with a harder schedule?
There is one easy thing I see from this—P5 is one cluster of difficulty, and G5 is another one. It’s two different leagues. (This is where people say “oh, duh”, but the blue-green and the red clusters could be squished together a lot more. Maybe it’s the conferences and all the in conference games, but there’s a real separation between the two “leagues”.

Then it gets into what you do with the data. Normally, you’d do the first chart. If you were only looking at P5 teams, you’d get a different result. There probably aren’t enough games played between the G5 and P5 teams to justify having one big regression between them (just a guess). I also wonder how good a metric “strength of schedule” is if we don’t have enough games between the G5 and P5 teams.

Also, Nate Manzo didn’t give a correlation coefficient or a hypothesis test. Without those, I kinda guess “meh, it’s probably not that strong a relationship” based on just my preconceptions. But, I can see how people would find the examples to justify their arguments of “schedule HARDER” vs “No! Schedule EASIER!”.
 

stinger78

Ramblin' Wreck
Messages
736
There is one easy thing I see from this—P5 is one cluster of difficulty, and G5 is another one. It’s two different leagues. (This is where people say “oh, duh”, but the blue-green and the red clusters could be squished together a lot more. Maybe it’s the conferences and all the in conference games, but there’s a real separation between the two “leagues”.

Then it gets into what you do with the data. Normally, you’d do the first chart. If you were only looking at P5 teams, you’d get a different result. There probably aren’t enough games played between the G5 and P5 teams to justify having one big regression between them (just a guess). I also wonder how good a metric “strength of schedule” is if we don’t have enough games between the G5 and P5 teams.

Also, Nate Manzo didn’t give a correlation coefficient or a hypothesis test. Without those, I kinda guess “meh, it’s probably not that strong a relationship” based on just my preconceptions. But, I can see how people would find the examples to justify their arguments of “schedule HARDER” vs “No! Schedule EASIER!”.
This is why we stratify data. Overall, it looks like the harder the schedule the better the resume'. However, when you stratify and look within the constraints of the two strata, you see both actually decline. I'd wager if he produced a GoF metric, the two latter charts will both produce better fits than the one composite chart.
 

roadkill

Ramblin' Wreck
Messages
839
There is one easy thing I see from this—P5 is one cluster of difficulty, and G5 is another one. It’s two different leagues. (This is where people say “oh, duh”, but the blue-green and the red clusters could be squished together a lot more. Maybe it’s the conferences and all the in conference games, but there’s a real separation between the two “leagues”.

Then it gets into what you do with the data. Normally, you’d do the first chart. If you were only looking at P5 teams, you’d get a different result. There probably aren’t enough games played between the G5 and P5 teams to justify having one big regression between them (just a guess). I also wonder how good a metric “strength of schedule” is if we don’t have enough games between the G5 and P5 teams.

Also, Nate Manzo didn’t give a correlation coefficient or a hypothesis test. Without those, I kinda guess “meh, it’s probably not that strong a relationship” based on just my preconceptions. But, I can see how people would find the examples to justify their arguments of “schedule HARDER” vs “No! Schedule EASIER!”.
Agree, which is why I've circled back to "inconclusive". Even if we isolate P5/G5, there doesn't appear to be a very strong correlation between the variables.
 

slugboy

Moderator
Staff member
Messages
10,100
On right now on CBS.


Iowa scored a TD. Time to lock in the over!

Edit: 5 minutes left and the under is still in play. Yeesh these OCs…

Edit 2: 1 minute left and no one can score. And…an interception. Overtime looks likely.

Edit 3: 16 seconds left and another int. Iowa in FG range with 7 seconds left. Virtually XP range. This game is drunk.

Edit 4: 13-10, FG as time expires Iowa wins, and Nebraska is 5-7 and short of bowl eligibility. The under hits.

Willy Wonka Suspense GIF
 
Last edited:
Top