All bow to the new BCS poll

[methomps]

The computers think Team A is better than Media darling? Hmmm... maybe media darling isn't really as great as everone thinks.

There have been six BCS bowl games where the computers disagreed with the human polls about which team was better. Five of the six times, the team that the computers had ranked higher lost. Media Darling 5, Team A 1.

If you tell me that the Trojans were a better team than LSU on a neutral field, fine. I can buy that. But that Sugar Bowl wasn't a neutral field, & nobody was beating the Tigers on that day.

I don't know. A one-dimensional passing team whose Heisman QB was turned into a liability by injury had 1st and 10 from midfield down by 7 with 2 minutes to go. I think the passage of time has cheated your memory of how close OU was to tying that game up late in the 4th quarter.

 

[sandgk]

The initial purpose of the computer rankings was to appease the AP poll. AP voters were worried (and it seems rightly) that being a part of the selection process would change their status from "reporting the news" to "making the news".

Which contention by the AP is the next most laughable example of parsing heard in recent years (the most laughable being in a political arena).

AP was always "making the news" when they reported the poll.

 

[methomps]

Which contention by the AP is the next most laughable example of parsing heard in recent years (the most laughable being in a political arena).

AP was always "making the news" when they reported the poll.

I think they drew the line at "reporting the poll" and "being 50% of what determined which teams go where for BCS bowl games"

 

[Sloopy45]

methomps: "I don't know. A one-dimensional passing team whose Heisman QB was turned into a liability by injury had 1st and 10 from midfield down by 7 with 2 minutes to go. I think the passage of time has cheated your memory of how close OU was to tying that game up late in the 4th quarter."

No, I remember the game. I just don't think any team in the country was walking into the Sugar Bowl that day & winning. I've got the utmost respect for that USC team, but I don't think they could've pulled it off. And I don't even think that LSU team was as good as the Press made them out to be.

Its a HUGE advantage to be one win away from a National Championship, playing that game on what is essentially your home field. Look at what USC did to scUM just days before within the friendly confines of the Rose Bowl. And that was an excellent scUM team.

 

[Buckeyeskickbuttocks]

There have been six BCS bowl games where the computers disagreed with the human polls about which team was better. Five of the six times, the team that the computers had ranked higher lost. Media Darling 5, Team A 1.
.

Thomps, that's an interesting factoid. I don't exactly follow what you mean, though. The team that made it in lost to the team they played?

If that's the case, then I don't think it supports your contention. That is to say, if a team makes it in over "media darling" that doesn't mean they are "better" then the team they play, just that they were "better" than the team they were ranked ahead of.

Additionally, while there is nothing but stats to go by, the data set is small and could just be an anomoly. You know what they say, any given saturday one team can beat another. Still, I'm not discounting the 5 - 1 on it's face, as again, we only have a few years upon which to evaluate so it's not your fault the data set is small.

 

[sandgk]

Thomps, that's an interesting factoid. I don't exactly follow what you mean, though. The team that made it in lost to the team they played?

If that's the case, then I don't think it supports your contention. That is to say, if a team makes it in over "media darling" that doesn't mean they are "better" then the team they play, just that they were "better" than the team they were ranked ahead of.

Additionally, while there is nothing but stats to go by, the data set is small and could just be an anomoly. You know what they say, any given saturday one team can beat another. Still, I'm not discounting the 5 - 1 on it's face, as again, we only have a few years upon which to evaluate so it's not your fault the data set is small.

Another perspective on that 5-1 stat. In all the years of the BCS there has been tweaking of the formula, adjustments to SOS and other "enhancements." As a result I think one can convincingly argue that there is no direct comparison between to 0-1 data point and any of the 1-0 years comprising the five media darlings sub-set.
When they have a track record of a statistically significant number of years (with 117 D 1A's say greater than 20 years) running an un-changed formula (for good or ill) then I would feel more comfortable analyzing the data.
The 5-1 Media Dolls vs Computer rankings would be considered censored data .. not part of this analysis.

Before the BCS reaches a total of 20 years with the same formula we are likely to see some form of play-off. At which point the debate would switch to who's hot versus who's not.

 

[methomps]

Thomps, that's an interesting factoid. I don't exactly follow what you mean, though. The team that made it in lost to the team they played?

If that's the case, then I don't think it supports your contention. That is to say, if a team makes it in over "media darling" that doesn't mean they are "better" then the team they play, just that they were "better" than the team they were ranked ahead of.

Additionally, while there is nothing but stats to go by, the data set is small and could just be an anomoly. You know what they say, any given saturday one team can beat another. Still, I'm not discounting the 5 - 1 on it's face, as again, we only have a few years upon which to evaluate so it's not your fault the data set is small.

I went and looked each year at the four BCS bowl games (Rose, Sugar, Fiesta, Orange). I then looked at the human and computer rankings for each of those teams. Most of the time, the computers and the humans both had one team ranked higher than the other. However, there were 6 games where the human polls had Team B ranked higher than Team A and the computer poll had Team A ranked higher than Team B. In those 6 games, the teams ranked higher by the human polls was 5-1.

It is a small data set, you are correct. I didn't want to expand out into regular bowls because I was more interested in how the polls did at ranking the best-of-the-best. I would've limited my search to just BCS Champ games, but then the data set would've been even smaller. I felt all BCS bowls was a good balance.

Another perspective on that 5-1 stat. In all the years of the BCS there has been tweaking of the formula, adjustments to SOS and other "enhancements." As a result I think one can convincingly argue that there is no direct comparison between to 0-1 data point and any of the 1-0 years comprising the five media darlings sub-set.
When they have a track record of a statistically significant number of years (with 117 D 1A's say greater than 20 years) running an un-changed formula (for good or ill) then I would feel more comfortable analyzing the data.
The 5-1 Media Dolls vs Computer rankings would be considered censored data .. not part of this analysis.

Before the BCS reaches a total of 20 years with the same formula we are likely to see some form of play-off. At which point the debate would switch to who's hot versus who's not.

Well, the 'study' wasn't really affected by changes in the BCS configuration because I wasn't looking at BCS rankings. I was looking at human polls (which have stayed consistent) and the computer poll. I suppose you could reset the counter when the NY Times computer left (I think that would make it 2-0 in favor of the computers), but you could also just ignore the NY Times poll even when it was a factor. That might add or subtract one game. I might do that if I have time.

As far as a 20-year study, either the BCS or the human element won't be around in 20 years (at least not in the same capacity as it is now). Even the AP poll leaving hurts the study because someone can claim "the AP voters no longer feel the burden of being a part of the system, so their voting methods and tendencies change".

 

[Buckeyeskickbuttocks]

Thanks for the explaination Thomps. I'm surprised by the 5 - 1.

Maybe humans are better than I give them credit for. Nah, can't be... if they were Entertainment Tonight would have been off the air after it's first episode.

I wonder, however, how the Computer Polls do against the rest of the top 25 - sounds like too much work for me to get in to, but it would be interesting to see how the puters stack up against the polls over all games. Truth is, I bet the results are pretty close either way.

The way I look at it, its more about 1 team being given a uneven shake (Cal last year is a good example - yes, I know Cal lost to TTU and Texas beat Michigan.. still, on the bowl selection day, Cal should have been in the Rose) rather than the system being flawed as a whole. That is, the "media darling" factor - where the media dubs what will be a 2 loss OU team "the greatest team of all time" - or USC got screwed (In 03) that's where the human polls are suspect (At least in my view)

 

[Sloopy45]

methomps: "However, there were 6 games where the human polls had Team B ranked higher than Team A and the computer poll had Team A ranked higher than Team B. In those 6 games, the teams ranked higher by the human polls was 5-1."

What were those six games & just what was the difference in these teams rankings? I think if you would post that, it would make your assertion much clearer, and would make your argument hold more water.

 

[BuckTwenty]

The human polls also helped boost Texas over Cal and into the Rose bowl vs. Scum last year, thanks to some embarressing whining from Mack Brown that they were going to be snubbed. I remember that there were signicant changes in vote, even though Cal did not play the final week and Texas did (for the Big 12 title game). I feel terrible for Cal, though this was one media darling vs. another media darling. Cal was unjustifyably held out of a bowl game they rightfully deserved by human pollsters.

I also vividly remember one of the human pollsters a few years ago not including the Buckeyes in the top 25 when they were clearly a top 5 team, which nearly swung them out of the national championship game, though not having a single loss. Why did he do this? In his own words, saying he had a bias against the Buckeyes.

Human polls add just what the BCS doesn't want in its formula: style points. It's a different form of style points than Spurrior or Bowden running up the score on a poor small 1-A school, but its style points all the same. Former players? You think you're going to get a non-biased result on former players?? You gotta be out of your mind.

These sort of stupid arguments will linger until the NCAA gets their collective heads out of their asses and imposes a playoff system. Then the arguments will cease, and the kids can go out and play some ball WITHOUT controversy.

 

[martinss01]

methomps, the only problem with your statistics is that the BCS isn't designed to pick who will win in the nc game. it is designed to place the two most worthy teams in a game together based upon their season. unlike a human who would compair two teams by stating o-line vrs. d-line, wr's vrs. dbacks, etc.. a computer bases its numbers 100% off of data fed into it from a teams record. that is why you can't state that the computer is less accurate than people in this fashion. the human poll is taking far more data into consideration and doing so in ways the computer isn't being permitted to.

there is an old saying in the pc world, "shit in, shit out". this is part of the bcs' problem. the pc isn't permitted to use all the available data. nor is it permitted to adequately measure the data it does get.

3 things need to happen. A. people need to accept that there will NEVER be a playoff and get over it. B. accept that the bcs CAN NOT prevent split nc's, just like every other selection process. there is no such thing as foolproof. C. the bcs needs alot more info than it is currently getting. remember, a pc is only as smart as the person who wrote its programming.

 

[methomps]

1999 Sugar

Ohio State (humans: 3 computers: 5.25)

Texas A&M (humans: 8.5 computers: 5.00)

Ohio State 24 Texas A&M 14

Humans 1 Computers 0

<o =""> </o>

2001 Orange

Oklahoma (humans: 1 computers: 1.86)

Florida State (humans: 3 computers: 1.29)

Oklahoma 13 Florida State 2

Humans 2 Computers 0

<o =""> </o>

2002 Fiesta

Oregon (humans: 2 computers: 4.83)

Colorado (humans: 3 computers: 4.50)

Oregon 38 Colorado 16

Humans 3 Computers 0

<o =""> </o>

2003 Orange

USC (humans: 5 computers: 3.67)

Iowa (humans: 3 computers: 4.83)

USC 38 Iowa 17

Humans 3 Computers 1

<o =""> </o>

2004 Sugar

LSU (humans: 2 computers: 1.83)

Oklahoma (humans: 3 computers: 1.17)

LSU 21 Oklahoma 14

Humans 4 Computers 1

<o =""> </o>

2005 Orange

USC (humans: 1 computers: 1.75)

Oklahoma (humans: 2 computers: 1.25)

USC 55 Oklahoma <!--[if !supportEmptyParas]--><!--[endif]--><!--[if !supportEmptyParas]--><!--[endif]--><!--[if !supportEmptyParas]--><!--[endif]--><!--[if !supportEmptyParas]--><!--[endif]--><!--[if !supportEmptyParas]--><!--[endif]--> 19

Humans 5 Computers 1

a computer bases its numbers 100% off of data fed into it from a teams record. that is why you can't state that the computer is less accurate than people in this fashion. the human poll is taking far more data into consideration and doing so in ways the computer isn't being permitted to.

On the contrary, that sounds like a decent argument for keeping the humans in the system.

 

[MililaniBuckeye]

methomps: Your above cherry-picking of BCS bowl games to suit your argument is laughable. There have been four BCS bowl games each year since the start of the BCS in the 1998 season, which means there have been 28 BCS bowl games, not the six you picked to support your argument. If you want to truly prove your point, do what you did above for all 28 bowl games and show us the results.

 

[methomps]

methomps: Your above cherry-picking of BCS bowl games to suit your argument is laughable. There have been four BCS bowl games each year since the start of the BCS in the 1998 season, which means there have been 28 BCS bowl games, not the six you picked to support your argument. If you want to truly prove your point, do what you did above for all 28 bowl games and show us the results.

Those were the only games where the humans and computers disagreed about who was the better team.

 

[sandgk]

Personally if find it deeply ironic that Oklahoma's score in that 2005 NC game is followed by plenty of <!--[if !supportEmptyParas]--><!--[endif]-->statements.