Stats Corner: Bias of the AP Poll

Bias can be defined by the Oxford Dictionary as “prejudice in favor of or against one thing or group compared to with another, usually in a way considered to be unfair.” Today on Stats Corner we are going to look at the bias in the AP football poll when comparing the losses of a ranked team to an unranked team between the Power 5, with 17 such losses, and Group of 5 teams, with 7 such losses. We are going to start in week 5 for two reasons 1) a Group of 5 ranked team did not lose to an unranked team until week 5 due to 2) the absurdity of having a preseason poll which is completely useless in determining which teams are the best in the country this year. This topic will be covered in a future Stats Corner at the end of the season.

Week 5

#3 Oregon lost at Stanford dropped 5 places

#10 Florida lost at Kentucky dropped 10 places

#15 Texas A&M lost to Mississippi dropped 12 places

#18 Fresno State lost at Hawaii dropped 17 places

#20 UCLA lost to Arizona State dropped 11 places

Average drop for home loss 11.5 places

Average drop for road loss 10.7 places, 7.5 place drop for Power 5, 17.0 drop for Group 5

Average drop for Power 5 9.5

Average drop for Group 5 17.0

Let us start with exhibit A in why the preseason poll is useless, Florida (as well as North Carolina, Washington, Virginia Tech, and USC) started the season ranked. At these point in the season, the polls were finally starting to clear out preseason mistakes, but Florida was still there. A road loss to Kentucky only dropped them 10 places, compared to a 17-point drop for the Bulldogs who had a much further road trip to the Islands, and Oregon’s trip to Stanford was not as long either. According to the polls, if you are a Power 5, losing at home is a lessor mistake than a G5 losing on the road in the middle of the Pacific Ocean.

Week 6

#1 Alabama lost at Texas A&M dropped 4 places

#10 BYU lost to Boise State dropped 9 places

For the time being, we are considering BYU a Group of 5, they can be a P5 when they join a conference. Alabama gets the benefit of the doubt being Alabama, they drop fewer places for a loss to an unranked opponent then most schools do losing to a ranked team (and a nice ranking in the selection committee ahead of some unbeaten teams. At least they ranked Oregon ahead of Ohio State).

Week 7

#2 Iowa lost to Purdue dropped 9 places

#17 Arkansas lost to Auburn dropped 11 places

#18 Arizona State lost at Utah dropped 8 places

#19 BYU lost at Baylor dropped 10 places

#20 Florida lost at LSU dropped 10 places

Average drop for home loss 9.0 places

Average drop for road loss 9.3 places, 9.0 place drop for Power 5, 10.0 drop for Group 5

Average drop for Power 5 9.0

Average drop for Group 5 10.0

The number two team loses at home to Purdue and only drops 9 places? While G5, for right now, BYU loses at current 7-1 and ranked Baylor and drops 10 places. Something smells fishy here.

Week 8

#7 Penn State lost to Illinois dropped 13 places

#8 Oklahoma State lost at Iowa State dropped 7 places

#14 Coastal Carolina lost at App State dropped 11 places

#18 NC State lost at Miami dropped 9 places

#25 Purdue lost to Wisconsin dropped 10 places

Average drop for home loss 11.5 places

Average drop for road loss 9.0 places, 8.0 place drop for Power 5, 11.0 drop for Group 5

Average drop for Power 5 9.8

Average drop for Group 5 11.0

Coastal Carolina has one blemish, on the road, and they almost drop out of the top 25, with 7 multiple lose teams ahead of them, including a 3 lose Penn State who lost at home to powerhouse 3-6 Illinois.

Week 9

#9 Iowa lost at Wisconsin dropped 10 places

#12 Kentucky lost at Mississippi St dropped 6 places

#17 Pittsburg lost to Miami dropped 8 places

#19 SMU lost at Houston dropped 4 places

#21 San Diego State lost to Fresno State dropped 11 places

#22 Iowa Sate lost at West Virginia

Average drop for home loss 9.5 places, 8.0 place drop for Power 5, 11.0 drop for Group 5

Average drop for road loss 8.3 places, 9.7 place drop for Power 5, 4.0 drop for Group 5

Average drop for Power 5 9.25

Average drop for Group 5 7.5

Mark Twain said there are “lies, damn lies, and statistics” and there is the common saying that the “exception makes the rules.” This is the exception week and the one week people would point to in order to show that there no P5 bias. SMU loss almost does not count as an unranked loss as Houston was just outside the top 25. This outlier means this is the only week where the Group of 5 dropped less than the P5. It is also the only week where multiple ranked G5 lost and just the second G5 home lost to an unranked team.

Overall Comparisons

Home loss 10.2 places dropped, 10.3 for P5 and 10.0 for G5

Road loss 8.9 places dropped 8.4 for P5 and 10.3 for G5

Power 5 schools dropped 9.1 spots for an unranked loss

Group 5 schools dropped 10.2 spots for an unranked loss

Obviously, a home loss is worse than a loss on the road, in terms of overall places dropped 10.2 compared to 8.9 a difference of 1.3 places or 14.6% differences. The differences between a home loss for a Power 5 school is comparable to a G5 school although the data set is shewed as there have been seven home losses for P5, but only two for G5. Unless you polling a married couple, two is a terrible size for a data set. Looking at the data, the AP considers a road loss for a G5 school comparable to a home loss to a P5 school, both drop the team by 10.3 places. When comparing road losses, a P5 school gets a definitive advantage 8.4 places compared to 10.3 for G5 a difference of 1.9 or 22.6%. The overall difference between the division is 1.1 places with an advantage towards P5 9.1 to G5’s 10.2 or 12.1% differences. In other words, a Power 5 school gets a 12% advantage when losing over a Group of 5 school. And the Power 5 should be glad that SMU lost otherwise, a road loss for a G5 would have 12.3 with a 46.4% difference and the overall loss for G5 is 11.4 with 23.9% difference. The outliner cut the differences in half to make it look less like the Power 5 have the huge bias that they do.

In short, the voters of the AP poll give the Power 5 conferences a bias of anywhere from 10-45% when looking strictly at losses to an unranked opponent. There is also a bias as it took until week 5 for the Group of 5 to even have a ranked team lose to an unranked team. Showing there is a bias starting with the preseason poll which makes it difficult for G5 teams to even get a shot at being ranked. But that is a future Stats Corner article.