Sagarin Pre Season Rankings

[UAalum72]

Sagarin's ratings have no credibilty as far as I'm concerned.

A few years ago Michigan was 2-1, Oregon was 3-0 including a win at Michigan, yet Sagarin had Michigan ranked ahead of Oregon. After that I became convinced that his system was seriously flawed.

I've heard Sagarin on the radio talk about this point. While looking at only two teams it looks bad, when you've got 200+ teams there are all kinds of circular defeat chains , because as you know on any given day..

Plus after three games his starting ratings still have value. Not until after five or six weeks do only current year results count.

[LeadBolt]

Any preseason ranking lacks validity. It takes at least 4 weeks to begin to get a feel. Sagarin is no different.

[89Hen]

Um, how does a computer ranking rank teams when there is no data to rank?

[BDKJMU]

Um, how does a computer ranking rank teams when there is no data to rank?

How does a poll rank teams when there is no data to rank?

Like I said before- preseason they're all equally worthless.

[BlueHen86]

Sagarin's ratings have no credibilty as far as I'm concerned.

A few years ago Michigan was 2-1, Oregon was 3-0 including a win at Michigan, yet Sagarin had Michigan ranked ahead of Oregon. After that I became convinced that his system was seriously flawed.

I've heard Sagarin on the radio talk about this point. While looking at only two teams it looks bad, when you've got 200+ teams there are all kinds of circular defeat chains , because as you know on any given day..

Plus after three games his starting ratings still have value. Not until after five or six weeks do only current year results count.

Then he shouldn't release his rankings until week 5 or 6.

[89Hen]

Um, how does a computer ranking rank teams when there is no data to rank?

How does a poll rank teams when there is no data to rank?

Like I said before- preseason they're all equally worthless.

Not even close to the same BDK. The computer rankings only rank wins, losses and scores. Human polls rank returning starters, new recruits, program tradition, etc... I'll take a human poll over the computers every day of the week and for pre-season, 8 days a week. Human polls usually have 15-20 of the final top 25 in the pre-season.

[clenz]

How does a poll rank teams when there is no data to rank?

Like I said before- preseason they're all equally worthless.

Not even close to the same BDK. The computer rankings only rank wins, losses and scores. Human polls rank returning starters, new recruits, program tradition, etc... I'll take a human poll over the computers every day of the week and for pre-season, 8 days a week. Human polls usually have 15-20 of the final top 25 in the pre-season.

There is a guy who does a computer rankings for football and boys basketbal in the state of Iowa for the high school level. He's some computer genious. He wrote the system himself, and in terms of ranking high school teams his system is BY FAR better than humans

[89Hen]

There is a guy who does a computer rankings for football and boys basketbal in the state of Iowa for the high school level. He's some computer genious. He wrote the system himself, and in terms of ranking high school teams his system is BY FAR better than humans

How so?

[clenz]

There is a guy who does a computer rankings for football and boys basketbal in the state of Iowa for the high school level. He's some computer genious. He wrote the system himself, and in terms of ranking high school teams his system is BY FAR better than humans

How so?

I don't understand how his system works, and i don't pretend too.

The long of it short is he has scores for every team (missing a couple games here and there from years ago) for the last 15-20 years. He takes past years success (tradition), SOS, etc... it typically has the state champion in the top 4 of the rankings by week 3 when they won't be ranked in human polls.

I'll find it and link it later

[LeadBolt]

I agree with BlueHen86, don't release the rankings until after week 4 or later.

[89Hen]

How so?

I don't understand how his system works, and i don't pretend too.

The long of it short is he has scores for every team (missing a couple games here and there from years ago) for the last 15-20 years. He takes past years success (tradition), SOS, etc... it typically has the state champion in the top 4 of the rankings by week 3 when they won't be ranked in human polls.

I'll find it and link it later

I don't need to know the method, just the results. Are you saying the state champion is usually a nobody?

[Col Hogan]

I agree with BlueHen86, don't release the rankings until after week 4 or later.

+1

[Skjellyfetti]

What magically happens at week 4 to make polls valid?

If you're going to get up in arms about polls... you should just wait until after the regular season is over for a poll... that's the only time you can legitimately measure teams, right?

[McNeese75]

Perhaps the most intersting thing about that listing is that Lamar, which is playing its first football game since the 1980s, is ranked 20th and higher than anybody in the Southland Conference it's going into. Not saying it's wrong as Lamar is a total unknown. I'm just wondering how that works...what Sagarin did to rate what's essentially a team from a brand new football program.

I see I'm not the only one who noticed.

As Sagarin's preseason ratings are supposed to be a weighted average of the past several season's final rankings, this point is very interesting. Lamar's rating of 57.36 is very close to the average of all teams FBS and FCS. So, it appears he is giving them an average rating to start.

Here is the reply to another Cowboy fan from Jeff Sagarin regarding the Lamar rating.

Good point; it will be fixed by next week.
My starting ratings routine looks over each team's
past history of ratings year-by-year, and since
Lamar had such ratings up through when they dropped
the sport, it used those ratings. I forgot to intervene manually.
The other two new teams, Georgia State and South Alabama
didn't have any rated seasons in my database and so I didn't
make that mistake with them.
Thanks for the good catch.
Jeff Sagarin

[ASUG8]

I'm not seeing the Richmond ranking making any sense......if they can stay in the top 15 with a new coach and the loss of most of their starters then great, but top 5 is a little hard to believe.

[clenz]

I don't understand how his system works, and i don't pretend too.

The long of it short is he has scores for every team (missing a couple games here and there from years ago) for the last 15-20 years. He takes past years success (tradition), SOS, etc... it typically has the state champion in the top 4 of the rankings by week 3 when they won't be ranked in human polls.

I'll find it and link it later

I don't need to know the method, just the results. Are you saying the state champion is usually a nobody?

The state champion isn't normally a nobody, typically there is 4 teams or so in every class you can mark down as top 10 teams almost every year. However, in the smaller classes teams that have been terrible can get a group of kids that come through that dominate but due to lack of traidition will never be rated. The perfect example of that is my high school basketball program.

For years we were a .500 team at best during a good year. We always had a ton of talent but a really shitty coaching staff. The class I was in lived sports every second of growing up. We started going to basketball tournaments in 1st grade, and it helped that out of the 5 kids in my class that stayed out for basketball all the way through our senior year 3 of them were 6'7 or taller, 2 of us could shoot 3's, the other kid...well he tried. Long story short my senior year of basketball as a team we broke pretty much every school record we could...points scored, least points allowed, wins, blocks, rebounds, steals, assits, most points scored in a game (125), largest margin of victory (95 final score was 125-30), largets average victory (35.4 points)first state tounrament appearance, etc... We never entered a top 10 poll. The highest we got was 14th thanks to the ORV. We get to district tournament time to qualify for state and win 3 of our 4 games by an average of 50 points, and beat 2 top 10 teams.

We get to state with a record of 21-4. We were the only team with more than one loss in the state tournament (we also played in a conference that had 4 of the ten teams qualify for the state tournament, and 6 teams that were ranked in the top 10 of their respective class). Another long story short, the first game of the tournament we beat the #3 team in the state by 10 but we stalled the entire second half because they wanted to sit in a zone when they were losing and didn't want to pressure the ball. Second game of the state tournament we play the #4 team in the state and beat them by 36. State title game we played a team tha that had been #1 all season long and had Jordan Eglseder on it (UNI's starting center that just graduated and is in camp with the Houston Rockets right now). We hold them to a state tourament record for least points scored in a quarter and win the state title.

Now, if someone looks at the polls that entire season (human polls) we should never have a chance against those teams, heck we should never have been there. However, this computer poll had us in the top 10 all season, and flip flopping between #1 and #2 after the first 10 games with the team we beat for the title.


The state champion doens't always finish #1 in this poll, but if they don't 98% of the time they are #2...the chances of that happening with the human polls that are put out is about 1 out of 100 unless there is a CLEAR #1 team. People that follow high school sports in Iowa that know about this poll often give this much more respect than the human polls. This ranking system also has the state qualifiers in his to 5-8 more often than the human polls do. It's quite the complex system (I don't know how Sagarin runs his so his may be as well) but it is more than the scores and SOS. Your opponents oppenent affects your ranking. The smallest school in Iowa ties into the largest school in Iowa the way his system works. It also plays into a predicted outcome vs actual outcome and blah blah blah...much better than human poll


http://www.bcmoorerankings.com" onclick="window.open(this.href);return false;

[89Hen]

My guess is the human pollsters don't really get enough Iowa HS basketball (or football) coverage. This doesn't hold true for Division I NCAA football. In I-AA, there is no computer ranking that has outperformed the AGS human poll.

[clenz]

My guess is the human pollsters don't really get enough Iowa HS basketball (or football) coverage. This doesn't hold true for Division I NCAA football. In I-AA, there is no computer ranking that has outperformed the AGS human poll.

The polls put out for Iowa High School sports is by Iowa writters and voters. The Des Moines Register, Radio Iowa Sports, etc...


I like the idea of what humans can bring into voting more than a computer, just like anyone else here. However, humans have something that can help them versus a computer...knowledge of the team past the scores, SOS, and whatever else goes into it. The knowledge of returning starters, returning stats, coaches, program tradition, etc...


However, that can also hurt them compared to the computer polls. The computer polls go "Team A based on the numbers should beat Team B" it doesn't matter if team A has been terrible and Team B has won a couple national titles. That bias based on team name isn't going to cloud the computers judgment.


I'm not saying I buy 100% into the computers, but I give them just as much credit as the human polls

[89Hen]

However, that can also hurt them compared to the computer polls. The computer polls go "Team A based on the numbers should beat Team B" it doesn't matter if team A has been terrible and Team B has won a couple national titles. That bias based on team name isn't going to cloud the computers judgment.


I'm not saying I buy 100% into the computers, but I give them just as much credit as the human polls

I don't. Who wrote the computer models? Humans. They're only as good as the programmer. They also can NEVER take into account what happened other than the scoreboard. A fluke turnover costs somebody the game. The game was played in a wind/rain storm. The stud RB got hurt on the first play.... computers only get the score. That's what makes them inferior. The only way human polls get it wrong is if the voters are biased or ignorant.

[clenz]

However, that can also hurt them compared to the computer polls. The computer polls go "Team A based on the numbers should beat Team B" it doesn't matter if team A has been terrible and Team B has won a couple national titles. That bias based on team name isn't going to cloud the computers judgment.


I'm not saying I buy 100% into the computers, but I give them just as much credit as the human polls

I don't. Who wrote the computer models? Humans. They're only as good as the programmer. They also can NEVER take into account what happened other than the scoreboard. A fluke turnover costs somebody the game. The game was played in a wind/rain storm. The stud RB got hurt on the first play.... computers only get the score. That's what makes them inferior. The only way human polls get it wrong is if the voters are biased or ignorant.

That's your choice then, and for the most part I agree with that sentiment.

However, lets say during a UD game this year Devlin goes down, and UD losses by 2 to West Chester (won't happen, but I'm just throwing this out there) That game will hurt UD in the computers, but the human pollsters will go, well Devlin was out that game doesn't really count. Again, for the most part I agree. However, you still should have beaten that team.


Now, lets say Devlin goes down for the season in that game. The human pollsters are still going to go, well they don't have Devlin that loss isn't that bad or that win was much more impressive than it really was. The computer doesn't care. A good computer set up goes beyond just you and your oppenents. That's why I like the one that bcmoore set up for the Iowa high school sports. It takes into account EVERY game. How your oppenenents oppenents oppenent does affects how your game against your oppenent does based on results built in on top of results.

With all of that said, I take all polls with a grain of salt. Like you said, computers don't take players/injuries/etc... into account, but humans tend to put too much weight on certain things with leads to some pretty big biases as well.

[89Hen]

A good computer set up goes beyond just you and your oppenents. That's why I like the one that bcmoore set up for the Iowa high school sports. It takes into account EVERY game. How your oppenenents oppenents oppenent does affects how your game against your oppenent does based on results built in on top of results.

With all of that said, I take all polls with a grain of salt. Like you said, computers don't take players/injuries/etc... into account, but humans tend to put too much weight on certain things with leads to some pretty big biases as well.

Stats are on my side on this one when it comes to DI football.

EVERY computer should take into account every score of common opponents. Are there really any that don't? However, the problem with football in particular is that you'd have a hard time finding a computer guy who can say with a straight face that there is enough data for them to be accurate. How many games will be played between the MVFC and the CAA this year? I know UD is playing SDSU. Are there any others? There is no way you can tell me a computer, whose ONLY data is scores, can make the leap to compare SIU to W&M.

Here are the CAA and MVFC entire OOC slates...

Delaware - West Chester, SDSU and Duquesne
JMU - Morehead, VT and Liberty
Maine - Albany, Monmouth and Syracuse
UMass - Holy Cross, Michigan and Stony Brook
UNH - CCSU, Pitt and Lehigh
URI - Buffalo, Fordham and Brown
Richmond - Virginia, Elon and Coastal
Towson - Indiana, Coastal and Columbia
Villanova - Temple, Lehigh and Penn
W&M - VMI, ODU and UNC

ISU - Central Missou, Northwestern and EIU
InSU - St Joseph's, Cincinnati and Quincy (good Lord BTW)
MSU - EKU, KSU and Murray
NDSU - Kansas and USD
UNI - SFA, IowaSt and USD
SDSU - Delaware, Nebraska and UND
SIU - Quincy, SEMO and Illinois
WIU - Valpo, Purdue, SHSU
YSU - PSU, Bulter, CCSU

So we have one game between the conferences and one game with a common opponent. That's it until playoffs. A computer, that can ONLY rely on scores, can rank WIU agains Maine?

[BDKJMU]

However, that can also hurt them compared to the computer polls. The computer polls go "Team A based on the numbers should beat Team B" it doesn't matter if team A has been terrible and Team B has won a couple national titles. That bias based on team name isn't going to cloud the computers judgment.


I'm not saying I buy 100% into the computers, but I give them just as much credit as the human polls

I don't. Who wrote the computer models? Humans. They're only as good as the programmer. They also can NEVER take into account what happened other than the scoreboard. A fluke turnover costs somebody the game. The game was played in a wind/rain storm. The stud RB got hurt on the first play.... computers only get the score. That's what makes them inferior. The only way human polls get it wrong is if the voters are biased or ignorant.

Thats the thing. Most of the voters are biased or ignorant for outside their region or conference. The computers have their flaws, so do the human polls. Preseason they're all as worthless as tits on a bull. By mid season they have something to go on, but they're still all somewhat flawed.

Whats that expression? Polls? We don't need no stinkin polls! We have the playoffs! (And you can substitute computer rankings for polls.

[89Hen]

Most of the voters are biased or ignorant for outside their region or conference.

Opinion.

[andy7171]

You guys do realize you are arguing with a native Iowegian, don't you? His thinking is flawless.