header pic

Perhaps the BEST B1G Forum anywhere, here at College Football Fan Site, CFB51!!!

The 'Old' CFN/Scout Crowd- Enjoy Civil discussion, game analytics, in depth player and coaching 'takes' and discussing topics surrounding the game. You can even have your own free board, all you have to do is ask!!!

Anyone is welcomed and encouraged to join our FREE site and to take part in our community- a community with you- the user, the fan, -and the person- will be protected from intrusive actions and with a clean place to interact.


Author

Topic: preseason polls coming out...

 (Read 13204 times)

bayareabadger

  • Legend
  • ****
  • Default Avatar
  • Posts: 7947
  • Liked:
Re: preseason polls coming out...
« Reply #84 on: August 07, 2019, 09:26:38 AM »
A computer poll preseason is bizarre to me.  What are the inputs?  Number of starters returning?

I guess they could take last year's finish and do something with starters returning, but or number of starts returning at each position, or something.

Quija board?
I know Bill Connelly's three inputs are recent success, returning production and recruiting rankings.

The first and last of those are pretty good predictors. The middle one he's spent a bunch of time modeling. I think they exist primarily because people hunger for rankings, but I know his have had some decent early success vs. Vegas lines. 

bayareabadger

  • Legend
  • ****
  • Default Avatar
  • Posts: 7947
  • Liked:
Re: preseason polls coming out...
« Reply #85 on: August 07, 2019, 09:27:22 AM »
if there's a new coaching staff does the computer factor the previous season at the other school?
I don't think so most of the time, primarily because coaching changes are so random. For every new staff of coordinator who saves the day, there's probably more who are just duds. 

Cincydawg

  • Oracle of Piedmont Park
  • Global Moderator
  • Hall of Fame
  • *****
  • Default Avatar
  • Posts: 73382
  • Oracle of Piedmont Park
  • Liked:
Re: preseason polls coming out...
« Reply #86 on: August 07, 2019, 09:29:54 AM »
A new OC or DC is unlikely to be great in year one unless his predecessor was really bad, I think.  Most will be vanilla relatively.  Some will be worse, though perhaps it is a result of injuries rather than play calling etc.

A new HC usually has a mediocre year followed by a very good year, from what I've seen.

utee94

  • Global Moderator
  • Hall of Fame
  • *****
  • Posts: 18082
  • Liked:
Re: preseason polls coming out...
« Reply #87 on: August 07, 2019, 10:46:15 AM »
Who are the computers polling?  Or what poll questions are being asked of the computers?

CWSooner

  • Team Captain
  • *******
  • Posts: 6109
  • Liked:
Re: preseason polls coming out...
« Reply #88 on: August 07, 2019, 11:37:47 AM »
A computer poll preseason is bizarre to me.  What are the inputs?  Number of starters returning?

I guess they could take last year's finish and do something with starters returning, but or number of starts returning at each position, or something.

Quija board?
Richard Billingsley--whose computer rankings were part of the BCS formula--explains it like this.

Quote
Starting Position- This is one of the most hotly debated subjects in rankings. Starting position DOES have an impact in rankings, but what I’ve done is create a system where the pre-season ranking is corrected early on (through strength of schedule and head to head results) so there is no undue advantage from the pre-season poll that remains past about the third week of September. I respect many different points of view here, ranging from creating a pre-season poll based on returning starters and media hype (as in the AP/Coaches), starting everyone equal (as in some computer polls), or having a starting position based on an average of 3-5 previous seasons (also in some computer models). I believe having a starting position is best, but starting everyone equal is not logical to me. We know through observation of past seasons that some teams are stronger than others. No disrespect to the Vandals, but in 2007 Idaho was not as strong a team as Texas. If we know this in advance, to a high degree of accuracy, then ranking Texas and Idaho equal is not only illogical, it is unfair to Texas and completely (in my mind) skews any hope of an accurate strength of schedule. In the past I kept teams in their earned rank positions from the end of one season to the beginning of the next. If a team finished #10 in 2007, they started #10 in 2008. For decades I felt it was the fairest way to establish a starting position. However, beginning in 2014 I began creating a pre-season ranking based on returning starters and coaching changes. It’s not so much that I changed my mind as it is the fact that enormous amounts of information are now available to me now through the internet, and being semi-retired, I have the time to devote to that process. I still adjust a team’s RATING to a standard point value that brings teams closer together, preventing an unfair advantage in points from one season to the next. Each year the #1 team starts at 270 points; #2, 269.250 points; #3, 268.500 points and so forth all the way to #128 starting at 174.150 points.

https://enc.cfrc.com/archives/ISO_08.htm

Play Like a Champion Today

Anonymous Coward

  • All Star
  • ******
  • Posts: 3187
  • Liked:
Re: preseason polls coming out...
« Reply #89 on: August 07, 2019, 12:06:22 PM »
I have no doubt that preseason starting position was an influencer (modest or more) in the BCS era. But Cincy has persuaded me that the effect has been lost in the CoFoPO era. I guess that's a function of their independence and their discipline in delaying their first ranking until the season is more than half completed. If they are any amount affected by the preseason polls, it's so much smaller than the BCS era phenomenon as to not register on our measuring stick.

medinabuckeye1

  • Legend
  • ****
  • Default Avatar
  • Posts: 9059
  • Liked:
Re: preseason polls coming out...
« Reply #90 on: August 07, 2019, 12:21:49 PM »
I have no doubt that preseason starting position was an influencer (modest or more) in the BCS era. But Cincy has persuaded me that the effect has been lost in the CoFoPO era. I guess that's a function of their independence and their discipline in delaying their first ranking until the season is more than half completed. If they are any amount affected by the preseason polls, it's so much smaller than the BCS era phenomenon as to not register on our measuring stick.
I agree entirely.  

The only thing I'll add is that the preseason ranking could theoretically impact the final rankings for teams #5-#25 because those get less scrutiny but I think it has zero impact on teams ranked #1-#4 in the final poll which is the only thing that REALLY matters in the CFP poll.  

I'm just assuming here that the committee spends a LOT more time on the #4 vs #5 decision than they do on the #14 vs #15 decision because everybody will debate and analyze the #4 vs #5 distinction while basically nobody outside of a few REALLY hardcore fans of team #15 will debate or analyze the #14 vs #15 decision.  

FearlessF

  • Hall of Fame
  • *****
  • Posts: 38515
  • Liked:
Re: preseason polls coming out...
« Reply #91 on: August 07, 2019, 12:30:07 PM »
I like the idear of starting everyone equal and not releasing a poll until mid November

it would be more fair imo but, that wouldn't be as much fun
"Courage; Generosity; Fairness; Honor; In these are the true awards of manly sport."

CWSooner

  • Team Captain
  • *******
  • Posts: 6109
  • Liked:
Re: preseason polls coming out...
« Reply #92 on: August 07, 2019, 12:50:40 PM »
I like the idear of starting everyone equal and not releasing a poll until mid November

it would be more fair imo but, that wouldn't be as much fun
Are fairness and correctness the same thing?
If they're not, I'd go with the system that ends up with the best team ranked #1.
In the passage from Billingsley's website I posted above, he makes the hypothetical case about Idaho and Texas.  We know with a high degree of certainty that Texas is better than Idaho, so what's the point of starting them ranked the same?
Play Like a Champion Today

medinabuckeye1

  • Legend
  • ****
  • Default Avatar
  • Posts: 9059
  • Liked:
Re: preseason polls coming out...
« Reply #93 on: August 07, 2019, 12:58:01 PM »
Are fairness and correctness the same thing?
If they're not, I'd go with the system that ends up with the best team ranked #1.
In the passage from Billingsley's website I posted above, he makes the hypothetical case about Idaho and Texas.  We know with a high degree of certainty that Texas is better than Idaho, so what's the point of starting them ranked the same?
I know that at least some (and I assume all) of the computer polls are set up such that preseason ranking diminishes as a factor each week until it is deleted entirely usually around or before mid-season.  

I agree with Billingsley's contention that we can predict with a high degree of certainty that Texas is better than Idaho and that consequently it is illogical to rank them as equals to start.  

That said, even though we can predict this with a "high degree of certainty" it doesn't always turn out to be true.  In a normal year there will be at least one team that started out pretty highly ranked that just flat out sucks and at least one team that started out unranked that turns out to be pretty good.  

If I ran a computer poll, here is what I would do:
  • Preseason poll: 100% based on prior year's rankings, recruiting rankings, returning starters, etc.  
  • Week 1 poll:  80% preseason, 20% results this season.  
  • Week 2 poll:  60% preseason, 40% results this season.  
  • Week 3 poll:  40% preseason, 60% results this season.  
  • Week 4 poll:  20% preseason, 80% results this season.  
  • Week 5 (and beyond) poll:  100% results this season.  
Thus, by mid-October the preseason bias would be eliminated.  


FearlessF

  • Hall of Fame
  • *****
  • Posts: 38515
  • Liked:
Re: preseason polls coming out...
« Reply #94 on: August 07, 2019, 01:04:06 PM »
Are fairness and correctness the same thing?
If they're not, I'd go with the system that ends up with the best team ranked #1.
In the passage from Billingsley's website I posted above, he makes the hypothetical case about Idaho and Texas.  We know with a high degree of certainty that Texas is better than Idaho, so what's the point of starting them ranked the same?
fairness and correctness may or might not be the same thing, probably not.

I'd rather be fair to Idaho.  Idaho and Texas each have 12 games to earn their ranking.  This will sort out after 12 games.

It's always been my opinion that the only poll that matters is the final poll.
"Courage; Generosity; Fairness; Honor; In these are the true awards of manly sport."

CWSooner

  • Team Captain
  • *******
  • Posts: 6109
  • Liked:
Re: preseason polls coming out...
« Reply #95 on: August 07, 2019, 01:14:05 PM »
I know that at least some (and I assume all) of the computer polls are set up such that preseason ranking diminishes as a factor each week until it is deleted entirely usually around or before mid-season. 

I agree with Billingsley's contention that we can predict with a high degree of certainty that Texas is better than Idaho and that consequently it is illogical to rank them as equals to start. 

That said, even though we can predict this with a "high degree of certainty" it doesn't always turn out to be true.  In a normal year there will be at least one team that started out pretty highly ranked that just flat out sucks and at least one team that started out unranked that turns out to be pretty good. 

If I ran a computer poll, here is what I would do:
  • Preseason poll: 100% based on prior year's rankings, recruiting rankings, returning starters, etc. 
  • Week 1 poll:  80% preseason, 20% results this season. 
  • Week 2 poll:  60% preseason, 40% results this season. 
  • Week 3 poll:  40% preseason, 60% results this season. 
  • Week 4 poll:  20% preseason, 80% results this season. 
  • Week 5 (and beyond) poll:  100% results this season. 
Thus, by mid-October the preseason bias would be eliminated. 
I think that's roughly how Billingsley does it.
Play Like a Champion Today

utee94

  • Global Moderator
  • Hall of Fame
  • *****
  • Posts: 18082
  • Liked:
Re: preseason polls coming out...
« Reply #96 on: August 07, 2019, 01:59:39 PM »
Do the computers conduct their polling over the internet?  Or are they embedded in some sort of humanoid robot body, to blend in better?

Is there any way to tell the humanoid robot computer pollsters from ordinary humans?  Are they out to destroy us?  Did Billingsley unwittingly create the first edition of SkyNet?

ELA

  • Global Moderator
  • Hall of Fame
  • *****
  • Posts: 20481
  • Liked:
Re: preseason polls coming out...
« Reply #97 on: August 07, 2019, 02:00:48 PM »
There's a week in October where seemingly all of those computers totally release the prior year's results as an input, because if you follow those Massey composites (as I post each week), there's always one week with a bunch of wonky movement, and it seems like it has to be due to that.

 

Support the Site!
Purchase of every item listed here DIRECTLY supports the site.