BobcatAttack.com - For Fans of Ohio University Sports

Welcome Guest!
Create an Account

site search

about us

advertise with us

help

BobcatAttack.com Message Board

Ohio Football

Topic: Week 5 Sagarin 62

Topic: Week 5 Sagarin 62

Author

Message

Bcat2
General User

Member Since: 7/6/2010
Post Count: 4,295

Status: Offline

Week 5 Sagarin 62

Posted: 9/30/2012 10:03:25 PM

http://usatoday30.usatoday.com/sports/sagarin/fbt12.htm

62

"Do not pray for easy lives. Pray to be stronger men." JFK

L.C.
General User

Member Since: 8/31/2005
Location: United States
Post Count: 10,497

Status: Offline

RE: Week 5 Sagarin 62

Posted: 9/30/2012 10:35:36 PM

Note that in the Predictor rating, which includes margin of victory, Ohio is #70, but in the Elo-Chess rating, used by the BCS, Ohio is #30.

�We have two ears and one mouth so that we can listen twice as much as we speak.� ― Epictetus

Nash'Cat
General User

Member Since: 3/29/2012
Location: Nashville, TN
Post Count: 156

Status: Offline

RE: Week 5 Sagarin 62

Posted: 9/30/2012 10:38:40 PM

L.C. wrote:

Note that in the Predictor rating, which includes margin of victory, Ohio is #70, but in the Elo-Chess rating, used by the BCS, Ohio is #30.

I did notice that. Keep winning, that's all the BCS cares about.

WE ARE OHIO
-Mike C.
Class of 2010
Media Arts & Studies - Music Production

The Situation
General User

Member Since: 7/12/2010
Location: Columbus, OH
Post Count: 957

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/1/2012 11:11:22 AM

I've been following the Sagarin's for a few years now. No matter what happens it feels like we're stuck in the 50-60 range. I honestly think they're just as ridiculous as anything else. (Possibly more ridiculous since the BCS rankings are actually giving us a respectable nod (30))

The Sagarin's have the C-USA ahead of the Sunbelt, WAC, and MAC. C-USA has a 9-32 out of conference record. Our MAC has a 21-22 record including 6 wins against the AQ. How is that reasonable under any circumstances? Conference USA's only win over an AQ school was a 25-24 Rice win over KANSAS!

Ohio's schedule rank right now is 157. I don't respect any man's equation which calculates 156 college football schedules to be more difficult right now than traveling to Penn State, Marshall, and UMass, while hosting NMSU, and Norfolk. There isn't a single 1-AA school that will have a tougher schedule than us all year. Let alone 30.

Last Edited: 10/1/2012 11:12:57 AM by The Situation

C Money
General User

Member Since: 8/28/2010
Post Count: 3,420

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/1/2012 11:15:50 AM

I don't think Sagarin is "well-connected" yet, and it's pretty useless until then.

oucs 1986
General User

Member Since: 12/20/2004
Location: Mason, OH
Post Count: 251

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/1/2012 3:54:36 PM

Check out Florida St's ranking, and their schedule strength.

The rankings are currently nonsense.

-john

Go Bobcats!

Victory
General User

Member Since: 3/10/2012
Post Count: 2,321

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/1/2012 9:06:08 PM

oucs 1986 wrote:

Check out Florida St's ranking, and their schedule strength.

The rankings are currently nonsense.

-john

I think Jeff Sagarin himself would say that right now they aren't worth much.

oucs 1986
General User

Member Since: 12/20/2004
Location: Mason, OH
Post Count: 251

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/1/2012 9:12:48 PM

Agreed.

-john

Go Bobcats!

L.C.
General User

Member Since: 8/31/2005
Location: United States
Post Count: 10,497

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/1/2012 9:30:51 PM

oucs 1986 wrote:

Check out Florida St's ranking, and their schedule strength.

The rankings are currently nonsense.

-john

I looked at Florida State's ratings, and I didn't see anything that looked unusual. What were you thinking was odd?

The ratings are getting close to being "well-connected", so I wouldn't call them nonsense. They may not be perfect, but they are much more reasonable now than a few weeks ago. Personally I have always considered his Elo-Chess ratings to be nonsense, and his Predictor rating to be his real rating. In the predictor rating, Ohio was originally rated close to 80, which was #30ish, but now it is down to 70, which is #70. Thanks to all the injuries, Ohio is not currently as good as they were at the beginning of the season. They are still good enough to be favored in all remaining games, but they will be favored by a lot less than they would be if they were healthy.

Last Edited: 10/1/2012 9:31:56 PM by L.C.

�We have two ears and one mouth so that we can listen twice as much as we speak.� ― Epictetus

oucs 1986
General User

Member Since: 12/20/2004
Location: Mason, OH
Post Count: 251

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/1/2012 9:48:30 PM

L.C. wrote:

oucs 1986 wrote:

Check out Florida St's ranking, and their schedule strength.

The rankings are currently nonsense.

-john

They're 8th, with a schedule strength of 140.

8 Florida State A = 89.10 5 0 57.71( 140) 0 0 | 1 0 | 89.69 10 | 88.78 7

Doesn't add up.

-john

Go Bobcats!

L.C.
General User

Member Since: 8/31/2005
Location: United States
Post Count: 10,497

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/2/2012 9:37:59 PM

The rating itself looks reasonable enough to me. The number I'd question is the strength of schedule being 140. The problem is probably the fact that the schedule includes some really bad teams, notably #180 Murray State, and #244 Savannah State, but it also includes a very good Clemson team. Beating #21, #180 and #244 is definitely harder than beating #147, #148, and #149, but the average would be the same. His rating system does account for this disparity, though you can't see it when you look at a raw number like "Strength of schedule".

Last Edited: 10/2/2012 9:39:40 PM by L.C.

�We have two ears and one mouth so that we can listen twice as much as we speak.� ― Epictetus

oucs 1986
General User

Member Since: 12/20/2004
Location: Mason, OH
Post Count: 251

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/2/2012 10:25:39 PM

agreed. but if you compare our SOS to theirs, they're almost 60 full places ahead of us... so clearly there's some historical data bias creeping in.

That's all I'm saying... this was the best example of the bias I could find.

-john

Go Bobcats!

L.C.
General User

Member Since: 8/31/2005
Location: United States
Post Count: 10,497

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/3/2012 12:28:52 AM

I don't think there is as much historical bias as you think. I believe they will be "well connected" after one more week, and I don't expect significant changes.

�We have two ears and one mouth so that we can listen twice as much as we speak.� ― Epictetus

oucs 1986
General User

Member Since: 12/20/2004
Location: Mason, OH
Post Count: 251

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/3/2012 1:15:47 PM

It's a bet!

-John

Go Bobcats!

Victory
General User

Member Since: 3/10/2012
Post Count: 2,321

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/3/2012 3:33:27 PM

I think that they will be well connected and that they will change. I have written my own systems and I can make good assumptions about how these work.

We would be ranked about #20 right now in the Sagarin-Elo if there wasn't bias. Our performance against UMass would be dragging us down enough that I still think we would be about #70 in the predictor. The first won't change to much with a win over Buffalo and we would get crushed in it if we lose. But we could get jumped or go up depending if teams below us beat good teams or teams above us lose. The latter is harder to predict because it is also depends on the score margin of the game.

Just a for instance, the Elo ratings rate you off of how hard your record was to achieve against your schedule. That's all it means really. It isn't a measure of who WOULD beat who. Oregon State is #11 in the Sagarin Elo but are undefeated against his #2 schedule. That doesn't make sense. Clearly, the preseason bias is pulling the Beavers down. Most "Advanced Retrodictors" as they sometimes say that don't have any preseaon bias right now have Oregon State #1 and since Sagarin's own system shows them unbeaten against the #2 schedule I think you'd see the same in his if they were unbiased.

The Situation
General User

Member Since: 7/12/2010
Location: Columbus, OH
Post Count: 957

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/3/2012 4:32:04 PM

Years ago I was under the impression the Sagarin's were some kind of unbiased ranking system. The more I look at it the more ridiculous it becomes. I don't know what "well connected" means, but by God I hope it's not the transitive property. There is no logical way to have an unbiased ranking system that has an initial rating for each team. I'm not really sure how to verbalize it, but even if a top rated team like #8 Arkansas (who's not that good) implodes and loses multiple times early (as they've shown), they are still distributing rating points to the teams on their schedule that were never theirs to begin with.

Anyways, look at last year's final, "well connected" Sagarins:

http://usatoday30.usatoday.com/sports/sagarin/fbt11.htm

North Dakota State finishes number 37 in the country! Absurd.
OHIO 10-3 (68) falls behind Sam Houston State (62) and Montana (63) who had three loses
Three loss Northern Iowa (75) and three loss Georgia Southern (81),
2-10 Kansas (91)!
12 additional 1AA schools in the top 120

How can this ranking system be respected with so many anomalies? What is he accounting for that allows so many FCS schools to move ahead of FBS schools? Because he ranks these FCS schools it unfairly and unrealistically punishes schools like OHIO, who instead of playing number 120 Akron, only gets credit for beating number 198! Akron.

L.C.
General User

Member Since: 8/31/2005
Location: United States
Post Count: 10,497

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/3/2012 5:16:26 PM

I personally tend to use the Sagarin "Predictor" because it uses margin of victory, which is a second piece of information, and it is information of value. If you only use "win-loss", and you only have 12 results for each team, and if you only feed in limited information, you are going to get a ranking of limited value. However, the BCS mandated that margin of victory could not be used by computers because they didn't want to encourage people to run up the score, so you get computer polls that are pretty silly. In human polls, humans are influenced by "style points", and don't just look at win-loss, so they are likely to be more accurate.

As far as the Sagarin Predictor, it simply reflects the games. The way I figure it, Ohio's game performances on the Predictor scale have been:
@Penn State +10 - so...take 77.70, add the 3 points for home, and the 10, and you get 90.70, a great game
NMSU +27 - so, take 50.52, -3 for it being a home game, plus 27, and you get 74.52, a decent game
@Marshall +3, take 64.69, add 3, and add 3, and you get 70.69, not quite as good
Norfolk State +34, take 40.87-3 +34, and you get 71.87, about the same
@U.Mass +3, take 45.27, add 3, and add 3, and you get 51.27, not a good score

Now, average 90.7,74.52,70.69,71.87, and 51.27 and you get 71.8, which is within error range of his ranking of 69.07, with the difference probably accounted for by the negative trend. It looks pretty reasonable to me, and you can do the same thing for any team.

As for the ELO-CHESS, well, Victory described how it works, and it's kind of a black-box. It would work a lot better for basketball or baseball where you have 30+ games than for football with only 12.

�We have two ears and one mouth so that we can listen twice as much as we speak.� ― Epictetus

Victory
General User

Member Since: 3/10/2012
Post Count: 2,321

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/4/2012 9:09:01 AM

L.C. wrote:

What you describe is really the simplest "Advanced Predictor" that you could write. Obviously, Our ranking is based off Penn State's, NMSU's, Marshall's, Norfolk State's and UMass's rating. Those schools ratings would be solved the same way. You'd have 124 equations in 124 unknowns. Its A LOT of math to solve that. It would take maybe decades or centuries to solve by hand but a computer can do it in a second. It is, however, very simple math. You learned how to do that in 7th grade. You desrcibed the simple equations to solve in your post above.

I have tried just what you said just to see in the past. Its not hard to write if you have any programming skills at all. After about 6 weeks you can look and see that you get a very similar result to Sagarin. That's almost certainly the basis for most of the Predictors. I'm sure there are little tweaks in there that their research shows thing on how to overweight more recent games, underweight total mismatches, etc. to get a more accurate predictor of future games.

As far as answering the question of "How hard was this record to post against this schedule?" there is going to be a little more variation. I don't know that it is exactly just a black box. But if you look at David Wilson's page you get different Classifications of the systems out there. In the BCS you have Sagarin-Elo, Massey BCS, Wolfe, and Colley classified as "Advanced Retrodictors" and in the BCS by the end of the they year give a very similar result. Its Billingsley (which is a more simple sequential system) and Anderson (which is more simple like the, IMO, awful RPI) that are more different in result.

Wes Colley's Algorithm is actually on his webpage. You can read exactly how it works and emulate it exactly if you want. If were the BCS and I wanted to use computer rankings then I would have insisted on this just like I would insist that every voter's ballot is public if I were going to use a poll.

Jeff Sagarin's algorithm is not available. However, the ELO name gives you a clue. That's the chess ranking system. You CAN look up exactly how that works. The difference is that the Chess system a sequential rating system per Wilson's classification system. In a sequential system last weeks ratings are modified off of the results of this weeks games and then you are done. We know that Sagarin's isn't sequential. So use the ELO system and instead of solving sequentially solve all ratings using all week's games at once (124 equations in 124 unknowns). You get very similar results to Sagarin's results once they are "well-connected" and unbiased.

Last Edited: 10/4/2012 10:09:06 AM by Victory

Ted Thompson
Administrator

Member Since: 11/11/2004
Location: MAC Play
Post Count: 7,720

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/4/2012 11:42:01 AM

The Situation wrote:

North Dakota State finishes number 37 in the country! Absurd.
OHIO 10-3 (68) falls behind Sam Houston State (62) and Montana (63) who had three loses
Three loss Northern Iowa (75) and three loss Georgia Southern (81),
2-10 Kansas (91)!
12 additional 1AA schools in the top 120

I too have scratched my head on the 1AA ratings. If it takes 6 games into the season to for 1A teams to be well-connected and there's so little inter-play between 1A and 1AA (certainly not 6 games worth on each schedule) then how can 1A and 1AA teams be ranked together on the same scale?

Further, there are some non-scholarship 1AA schools like Dayton that NEVER play 1A schools. Yet, according to the 2011 rankings, Akron was ranked 198th with a predictor of 40.75 and Dayton was 210th with a predictor of 40.73. This would say a team with 85 scholarship players and a team with none are equal on a neutral field. But if 1AA non-scholarship schools never play 1A schools how is this validated?

I actually sent an e-mail to Sagarin a couple of years ago asking him about this. He replied back with "Oh yeah, Appalachian St. could never beat Michigan either." Didn't really answer what I thought was a politely worded and fair question.

Follow Ohio Football recruiting on the BobcatAttack.com football recruiting database.

Victory
General User

Member Since: 3/10/2012
Post Count: 2,321

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/4/2012 11:49:49 AM

I think the connectivity between 1A and 1AA would be a problem. Probably less so on the Predictive rating side.

Anyway, I don't think that Akron last year was #194 either. But, lets not be foolish and call them #120. If we dimiss 1AA teams just because they are 1AA then we do the same thing that AQs do to non-AQs. They were bunches and bunches of 1AA teams that would have been favored to beat Akron last year.

Ted Thompson
Administrator

Member Since: 11/11/2004
Location: MAC Play
Post Count: 7,720

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/4/2012 12:05:51 PM

Victory wrote:

I agree. Akron was horrible. I'm just not sure they were even from a predictor standpoint with non-scholarship Dayton.

Follow Ohio Football recruiting on the BobcatAttack.com football recruiting database.

L.C.
General User

Member Since: 8/31/2005
Location: United States
Post Count: 10,497

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/4/2012 4:10:28 PM

I don't see a problem on the Predictor side, even with a team like Dayton. While they don't play any Division I teams, they do play other teams with Predictor rankings, so it's perfectly possible to compute a predictor ranking for them. One clue that the Predictor rankings aren't too off the mark that you can use the Predictor ranking to compute a point spread, and it is usually pretty close to what Las Vegas comes up with.

I do agree, however, that there is no way there is enough connectivity to use ELO-Chess to rank FBS and FCS schools on the same index.

�We have two ears and one mouth so that we can listen twice as much as we speak.� ― Epictetus

L.C.
General User

Member Since: 8/31/2005
Location: United States
Post Count: 10,497

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/7/2012 11:48:53 AM

L.C. wrote:

I don't think there is as much historical bias as you think. I believe they will be "well connected" after one more week, and I don't expect significant changes.

The teams are now well connected, and there is no significant change. The current ratings show that Ohio dropped from 30 to 35 in Elo-Chess, and from 72 to 76 in predictor, where the rating fell from 69.9 to 68.6. Ohio now trails both NIU and Toledo in Predictor by a small amount.

Realistically, with as many players injured as they have, Ohio is not nearly the team they were at the beginning of the season. The predictor reflects that better than the Elo-Chess ratings.

Last Edited: 10/7/2012 11:50:30 AM by L.C.

�We have two ears and one mouth so that we can listen twice as much as we speak.� ― Epictetus

Mike Johnson
General User

Member Since: 11/11/2004
Location: North Canton, OH
Post Count: 1,736

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/7/2012 12:13:31 PM

L.C. wrote:

I don't think there is as much historical bias as you think. I believe they will be "well connected" after one more week, and I don't expect significant changes.

The teams are now well connected, and there is no significant change. The current ratings show that Ohio dropped from 30 to 35 in Elo-Chess, and from 72 to 76 in predictor, where the rating fell from 69.9 to 68.6. Ohio now trails both NIU and Toledo in Predictor by a small amount.

Realistically, with as many players injured as they have, Ohio is not nearly the team they were at the beginning of the season. The predictor reflects that better than the Elo-Chess ratings.

How many teams are? It seems that on most days the football news includes reports of players injured and out for anywhere from a game to the remainder of the season.

Going into yesterday's game, we know that Buffalo was without starting RB Brandon Oliver. Any others?

Akron, for example, lost a starting DL yesterday - after already having lost another for the season. And so on through the ranks of the FBS teams...

Next Saturday in Athens I will have as guests a Dutch couple who will be seeing their first American football game. I'm anticipating that one of their questions will be: Why so many players on each team.? Part of my answer will be the frequent occurence of injuries.

http://www.facebook.com/mikejohnson.author

ord33
General User

Member Since: 12/30/2005
Location: Sevierville, TN
Post Count: 45

Status: Offline

RE: Week 5 Sagarin 62

Posted: 10/7/2012 1:57:33 PM

It still amazes me that North Dakota State is ranked #37 with a 185th rated schedule. I know they are 5-0 and have some decent wins like Colorado State, Northern Iowa, and Youngstown State, but I think #37 is definitely a stretch!

Northern Iowa supposedly has the nations 13 toughest schedule? They are 1-4 against that schedule with their ONLY win being against Central State and they are #95. Ahead of 9 MAC Teams. The only MAC teams higher are NIU, Ohio, Toledo, and WMU. For a team with their ONLY win against Central State that is absurd! Central State is 2-4 with their wins coming against Urbana and William Jewel College. They lost to Missouri University of Science and Technology, Saint Joseph's College, and Stillman College. Basically Northern Iowa has no quality wins and they are #95 and 13th toughest schedule.

Maybe the computers don't lie, but common sense tells me they aren't always accurate and tell the whole picture.

Last Edited: 10/7/2012 2:03:12 PM by ord33

Showing Replies:	1 - 25 of 28 Posts
Jump to Page:	1 \| 2 Next >

View Other 'Ohio Football' Topics

There are 316 Active Visitors on BobcatAttack.com

313 Guests
3 Registered Users:
	OhioCatFan, cbarber357, bobcatsquared

There are no other users looking at this page.