Pollspeak: Humans vs. computers

2008-11-19T16:40:31-05:00

2008-11-17T14:49:28-05:00

2008-11-18T21:48:15-05:00

http://img.fannation.com/upload/user_profile/image/581/405/thumb/CF5TAAROW-SD_1_.jpg 1338 RI

Armpit again

Portsmouth 581965

Ceteris Paribus between Texas and OU, the head-to-head has to be the ultimate tie-breaker. That said, Texas advances to the Big 12 Title Game. Any other result should result in the pollsters going directly to the whipping block - Do not pass Go, Do not Collect $200.00.

18 I think what is missing here is it is not a two team race. TT, TX, and OU would all have only one loss. The fairest way then would be to go with SOS, which would favor OU over TX as long as CINCY and TCU win out. 3378741

2008-11-19T16:40:30-05:00

http://img.fannation.com/upload/user_profile/image/604/839/thumb/100px-Utah_Utes_logo_1_.png 429 UT

UTEBROOK

Salt Lake City 605479

I think what is missing here is it is not a two team race. TT, TX, and OU would all have only one loss. The fairest way then would be to go with SOS, which would favor OU over TX as long as CINCY and TCU win out.

19 Although it is unclear which of these 3 teams is the best of the Big XII South, it seems to me that, if TTech loses to OU, then Texas should clearly get the conference Championship birth. Why? Because Texas and OU was a neutral field and Texas won. TTech beat Texas at home and OU would have beaten TTech at home. The only game not won by the home team is Texas/OU. Since both teams have evenly represented fan support at that game, it only seems fair to conclude that Texas/OU had the most even playing field and Texas won. 3388801

2008-11-18T18:34:21-05:00

Scott in Dallas

Flower Mound 546678

17 Ceteris Paribus between Texas and OU, the head-to-head has to be the ultimate tie-breaker. That said, Texas advances to the Big 12 Title Game. Any other result should result in the pollsters going directly to the whipping block - Do not pass Go, Do not Collect $200.00. 3377392

2008-11-18T21:48:15-05:00

http://img.fannation.com/upload/user_profile/image/581/405/thumb/CF5TAAROW-SD_1_.jpg 1338 RI

Armpit again

Portsmouth 581965

2008-11-18T18:34:21-05:00

Scott in Dallas

Flower Mound 546678

2008-11-18T16:26:40-05:00

GlennTN

Nashville 578339

16 A more in-depth article could have really gotten to the heart of the problems with many of the computer ranking systems. Most simply look at wins and losses and weight all games equally, regardless of when in the season they are played and whether or not they are conference games (important for qualifying for the BCS). Billingsley isn't so limited in his system. I like a lot of the thought that went into Billingsley's system, because I think there need to be some more factors built in, but he errors in what they are. While I think it's reasonable to have initial bias based on last year's results, it needs to either go away over time or be minimized. He does neither. He mentions the great example of Oregon losing games after Dixon got hurt and how that didn't affect the rankings of the teams that had already played Oregon. However, teams that played Kansas, Hawaii, Tennessee, or Michigan early in the season got unfairly rewarded based on how well those teams did last year. There are probably many better solutions like groupings of 10-20 teams and using average ratings, and/or recalculating the initial ratings based on the ratings after a few weeks. He rewards teams for defense, but irrespective of how good a team's offense is. If a team averaging 20 points a game is held to 3, that's more impressive than shutting out a team averaging 10 points. Also, an undefeated team should have more points by virtue of having won more games, not by being given a bonus for being undefeated. Ohio State is unfairly punished for having played a tough opponent early and losing. 3375652

2008-11-18T15:25:59-05:00

spectra

Cedar Creek 610754

14 Klee Irwin - You are right on. Texas beat OU head to head, end of story. If OU overtakes Texas in the BCS after beating TT, it will be purely political. Texas wins the computer stats in that part of the BCS formula. Unfortunately, they have to deal with the human polls that have short memories, and a love affair with OU and Florida. Florida is a fine team this year, but would not fare well against Texas or Texas Tech, or apparently, Ole Miss., either. 3374472

2008-11-18T16:01:05-05:00

204

GlennAllen

Richmond 185259

If OU overtakes Texas in the BCS after beating TT, it will be purely political.

15 Since when is the BCS (and all polls) anything but? 3375175

2008-11-18T15:25:59-05:00

spectra

Cedar Creek 610754

2008-11-18T13:06:27-05:00

123

PaulAnderson

Houston 610779

11 zlh67 - How is UT's strength of schedule better than OU's? By the end of the season, both teams will have played each other, Kansas, Tech, and OSU. Yes, Texas also played Missouri (#12) this year, but Oklahoma played Cincinnati (#19) and TCU (#15). Strength of schedule goes to OU. Also, you're forgetting that OU had a key injury in the Texas game, and they were ahead for most of the game (as was Tech). Combine all of that with the fact that OU's loss will be the earlies of the three (assuming they beat TTU and OSU), and I think they have a great argument for winning the 3-way tie. 3372143

2008-11-18T15:05:37-05:00

1595

Klee Irwin

Nome 582256

zlh67 - How is UT's strength of schedule better than OU's? By the end of the season, both teams will have played each other, Kansas, Tech, and OSU. Yes, Texas also played Missouri (#12) this year, but Oklahoma played Cincinnati (#19) and TCU (#15). Strength of schedule goes to OU. Also, you're forgetting that OU had a key injury in the Texas game, and they were ahead for most of the game (as was Tech). Combine all of that with the fact that OU's loss will be the earlies of the three (assuming they beat TTU and OSU), and I think they have a great argument for winning the 3-way tie.

13 it boils down to is beating TCU and Cincinnati at home, along with possibly beating Texas Tech at home, outwiegh losing head to head to Texas and the fact that Texas played 4 ranked teams in a row (including a ranked conference team Okla didn't play) and lost to TTech on the last play of the game on the road? Its a very close question, but head to head is hard to overlook and should be the ultimate tiebreaker. 3374124

2008-11-18T12:57:59-05:00

2062

Sandy Underpants

Los Angeles 481510

10 There's no thread on Mandel's projected Bowl Games, but if those come to fruition, what a sad end to such an exciting season. Texas vs. Ohio St? USC vs. Penn St? Alabama vs. Utah and Maryland(!) vs. Cincinnati-- this is truly a sad match-up that really lacks the ring of BCS bowl. Hopefully it ends up with a more exciting mix like USC vs. Texas/Alabama. I'd like to see Penn St. vs. Utah/Oklahoma. Texas vs. Alabama. Texas Tech vs. USC/Alabama. There should be a rejection option for stinkers like Maryland and Cincy, and won't it be Miami who goes to the Orange Bowl anyway? Miami vs. West Virginia sounds a heck of a lot better than what Stewey has. 3372022

2008-11-18T14:38:03-05:00

1595

Klee Irwin

Nome 582256

There's no thread on Mandel's projected Bowl Games, but if those come to fruition, what a sad end to such an exciting season. Texas vs. Ohio St? USC vs. Penn St? Alabama vs. Utah and Maryland(!) vs. Cincinnati-- this is truly a sad match-up that really lacks the ring of BCS bowl. Hopefully it ends up with a more exciting mix like USC vs. Texas/Alabama. I'd like to see Penn St. vs. Utah/Oklahoma. Texas vs. Alabama. Texas Tech vs. USC/Alabama. There should be a rejection option for stinkers like Maryland and Cincy, and won't it be Miami who goes to the Orange Bowl anyway? Miami vs. West Virginia sounds a heck of a lot better than what Stewey has.

12 that's what happens when you give automatic bids to BCS conference winners. So the ACC and Big East champs could have 3 losses and go to a BCS game. The Big East winner could be unranked. The lame Pac 10 may get TWO teams in the BCS with arguable the 7th or 8th best conference in America. That's what you get when you have the money grubbing BCS conferences running the show. Their should be a rule for qualifying for BCS bowls similar to what Notre Dame has----finish in the top 8 for an automatic bid. Proof that the system is rigged for the money grubbing BCS conferences, of which USC is a member. 3373570

2008-11-18T13:06:27-05:00

123

PaulAnderson

Houston 610779

2008-11-18T12:57:59-05:00

2062

Sandy Underpants

Los Angeles 481510

<div class="photo_container image_right"> <div class="photo"><img src="http://sportsillustrated.cnn.com/2008/images/11/17/unc-maryland-swp.jpg" alt="Double Click to select a Photo" /></div> <div class="photo_attributes" style="text-align:right;">Only one computer put Maryland in the top 25 after it beat UNC.</div> <div class="photo_attributes" style="text-align:right;">AP<br /></div> </div> <p><em>By Hugh Falk, <a href="http://www.pollspeak.com/" target="_blank">Pollspeak.com</a></em></p> <p>While the <a href="http://pollspeak.com/pollstalker/pollstalker.php?s=5&p=8&t1=0&t2=0&v=403&w=13&r=V" target="_blank">Billingsley Report</a> is one of the six BCS computer components, it sometimes resembles the human polls more than its computer brethren. This intentionally reflects <strong>Richard Billingsley</strong>'s desire to more closely mimic the way humans rank football teams. For example, this ranking prominently features rules for head-to-head results. While the effects of the head-to-head results only last one week, this is a major differentiator for the Billingsley Report. This week, we can see the difference between this system and other computer poll methodologies.</p> <p>Billingsley is the only computer to rank Oregon and Maryland in the top 25. It is also the only computer to not rank North Carolina and Miami. However, this is very similar to how the human BCS components voted. The only exception is North Carolina, which split the human vote -- Harris Interactive ranked the Tar Heels in the top 25, but the coaches didn't. Also, Billingsley agrees with the human polls on who should be No. 1 (Alabama) and No. 2 (Texas Tech). No other computer poll cast the lone vote for a top 25 team, yet Billingsley did so for four teams.</p> <p>I've heard concerns with Billingsley's formula in the past. It initially ranks teams based on how they finished the previous season. This seemingly gives Billingsley's rankings a type of "preseason bias," a criticism that, once again, sounds very human. Billingsley's methodology document is unique and a surprisingly easy read (relative to other computer methodologies). If you want to know more, you can <a href="http://www.cfrc.com/Archives/Dynamics_08.htm" target="_blank">read it here</a>. Overall, I like Billingsley's methods; however, preseason bias is a human fault that would be good for a computer ranking to avoid.</p> <p>Speaking of human faults, <a href="http://pollspeak.com/pollstalker/pollstalker.php?s=5&p=9&t1=0&t2=0&v=398&w=13&r=V" target="_blank">Maurice Patton</a> is this week's voter in the spotlight. <a href="http://pollspeak.com/pollstalker/pollstalker.php?s=5&p=9&t1=0&t2=0&v=0&w=13&r=F&o2=votes" target="_blank">Pollstalker</a> called out the <em>Tennessean</em> scribe as the voter with the most "extreme" votes (he had eight) this week. Here are some head-to-head results he's ignoring:</p> <p>-    Oklahoma (9-1) over Texas (10-1)<br />-    Missouri (9-2) over Oklahoma State (9-2)<br />-    North Carolina (7-3) over Maryland (7-3)</p> <p>In each case, he ranks the loser over the winner even though the loser has the same (or worse) record. Additionally, in each case the gap between the two teams is pretty wide (four or five spots). If only Patton had a head-to-head rule like Billingsley. Even if it only lasted one week, at least Maryland's win over North Carolina last Saturday would be covered.</p> <p>Finally, <strong>Barack Obama</strong> has once again publically endorsed a college football playoff, this time on <em>60 Minutes</em>. The president elect said he may "throw his weight around" and try to make an eight-game playoff a reality. Once he gives that a go, dealing with Congress should be a piece of cake.</p> <p>For more poll analysis, go to <a href="http://www.pollspeak.com/" target="_blank">pollspeak.com</a>.</p> 24084

The SI Staff

New York City 21244