Breaking Bias in College Football

A Grassroots Effort for a Level Playing Field

Rankings and Strength of Schedule

Posted by ncaafbsfootball on March 24, 2009

Every week during the football season various polls are released ranking the best teams.  The term best, meanwhile, is a vague, intangible superlative.  Sometimes the rankings reflect who would likely win if the teams were pitted against one another, and sometimes the rankings reflect consensus of which teams are more deserving of opportunity and prestige.  Whatever the case may be, all subective rankings appear to be a function of pre-season assumtions and wins and losses, rather than a fresh ranking each week.  Thus, losing teams almost always drop, but non-losing teams often move up only when teams ahead of them lose.

This relationship brings about an assortment of odd instances.  For instance, a team may win one week and drop a few spots in the polls only to move up even more spots the next week despite having a bye.  Other times, a slightly lower-ranked team can defeat a slightly higher-ranked team, yet remain ranked lower when the new polls are released.  Furthermore, because teams with greater notoriety are more likely to begin the year ranked highly in the polls, a bias exists favoring such teams.  Such inconsistencies call into question the worth of polls.

Strength of Schedule (SOS) is typically a calculated measure of the average difficulty of a team’s scheduled games.  For several important reasons, these measures must be taken with a grain of salt.  First, SOS is based on teams’ rankings, which are influenced by SOS ratings.  This circular logic, while it can be objectively automated, ultimately must be based on a beginning rating, which almost inevitably introduces and entrenches bias in the measure.  Second, any given teams’ rating will fall after losing, so by winning a team effectively lowers its SOS rating.  While this is true for all winning teams, the degree of change may not be the same for all teams.  Because teams unfavored by the inherent bias generally play more teams also unfavored by the bias, there is reason to speculate that the negative effect on an opponents SOS is greater in magnitude for already unfavored teams.  Third, SOS implicitly ignores the ups and downs teams naturally experience, and cannot accommodate the fact that teams match up differently against some than others.  Fourth, SOS factors in the average difficulty of all games, and thus creates bias against teams in conferences perceived as weak.  It’s true that it is more difficult to play tough teams week in and week out, so playing a single good team may not reliably serve as a litmus test of a team’s strength; however, it can be argued that a team playing a few good teams and a few more poor teams has a tougher road than a team that plays a straight set of average teams.  Because of this fact, no matter how many top teams a non-AQ team schedules in non-conference play it will be at a perhaps unjustifiable disadvantage because of SOS ratings.

Advertisements

Posted in Uncategorized | Tagged: , , | Leave a Comment »

Scheduling Maintains Bias in College Football

Posted by ncaafbsfootball on March 24, 2009

This report summarizes an original statistical analysis* performed with the goal of objectively rating the eleven FBS conferences and independents, and discusses the role of scheduling in college football.

Home Field Advantage – Over the past five years FBS teams have faced one another in 3,455 football games with the home team winning 59.3% of the time.  The statistical probability of that occurrence by chance alone is 1.84 x 10-26 (about 500,000 trillion times less likely than winning the lottery).   The question of practical significance may be answered by the fact that this 59.3% winning percentage means a school has a 50% greater chance of winning when playing at home.

Not all FBS schools play an even number of home and away games against other FBS schools.  Strong financial incentives drive games to be scheduled at locations that can seat more fans, resulting in more home games for the bigger schools.  However, this relationship only accounts for a portion of a team’s ability to schedule home games.  Reputation, most notably a team’s belonging to one of the six elite “BCS” conferences, also plays a role in scheduling.
Chart

As the chart above illustrates, there is a positive relationship between a school’s average attendance and its ability to schedule home games against BCS teams, meaning that schools with smaller fan bases are at a scheduling disadvantage.   Interestingly, however, this relationship does not hold true to the same degree when considering BCS and non-AQ schools separately.  As can be seen above, even within a specified range of average attendance representing multiple BCS and non-AQ schools, the BCS schools have generally higher home ratios, indicating that a degree of additional scheduling clout comes from merely belonging to a BCS conference.

Objective Measurement of Teams and Conferences – Eliminating the factor of home-field advantage reveals that while the BCS conferences are truly better than the non-AQ conferences, that superiority is exaggerated, and cannot serve as an objective basis for allocating funds and opportunities, especially in the case of the Mountain West.

Two methods may be used to adjust for this advantage: calculating wins per home game and variance from expected win ratio.  Both measures consider only opponents from BCS conferences.  Therefore, the average wins per home game ratio for all BCS conferences should be 1.00 since every game between two BCS schools has both a home and away team and a winner and a loser.  This creates a standard of comparison for all conferences with 1.00 being the average strength of all BCS conferences.  The table below shows tight variation between the six BCS conferences, Notre Dame, and the Mountain West with large discrepancies thereafter.
Table

A conference’s expected win ratio can be calculated simply by multiplying its ratio of home games by the overall FBS home team winning percentage (59.3%) and adding it to the away ratio multiplied by the FBS away team winning percentage (100% – 59.3%).  This approach inherently weights all conferences and teams with the same skill.  In reality, however, some conferences and teams are better than others, resulting in wildly varying winning percentages.  By comparing the expected winning percentage to the actual record, we can judge how close to average a particular team or conference is.  By way of context, a variance of zero would imply that particular conference is as strong as the average BCS conference.  As a result we expect to see some BCS conferences above and below zero, and all non-AQ conferences below zero.

Similar to wins per home game analysis, variance analysis shows a relatively tight grouping of the six BCS conferences, Notre Dame, and the Mountain West around the standard.  While the Mountain West may be deemed inferior to the average BCS conference by these analyses, such inference ignores the important fact that the natural break in relative strength comes after the Mountain West, indicating that the Mountain West is close being on par with the BCS conferences.  This relationship shows that the degree of superiority of the BCS over the Mountain West assumed by many is grossly overstated, so much so, that performance measures may not be reasonably relied on support the inequitable allocation of resources and opportunity facilitated by the BCS.  Yet, despite having much more in common with the BCS conferences than the other non-AQ conferences, the Mountain West remains a non-auto-qualifier.

picture-3

While the Rocky Mountain region remains sparcely populated compared to the rest of the US, it contains a number of large population centers currently excluded from the BCS.

Future Growth of the MWC – The similarity between the Mountain West and BCS is not surprising considering the history behind the six BCS conferences coming to power.  When the alliances between the big conferences and the high-paying bowls evolved into shape, the Rocky Mountain region was still sparsely populated.  With such small resources and recruiting bases, the schools in that region didn’t compare to the future BCS schools in more populous locations (see Map below).

Since 1990 the Rocky Mountain States have shown the fastest growing populations in the US, most of which growth has occurred in the metropolitan areas.  Though the populations of the Rocky Mountain States are still smaller than the rest of the country, it might be argued that they have caught up enough to be comparatively competitive.

Furthermore, the population increases of other states, most notably Texas and California, which border the rocky mountain states, combined with the NCAA’s limit on scholarships adds up to a growing number of top-caliber football players for roughly the same number of teams.  Many of those players are finding homes with Mountain West schools.

As the Rocky Mountain region continues to grow, the Mountain West will likely continue to improve.  It can be argued that the conference has already arrived at the doorstep of the big conferences, and due recognition of that accomplishment is now just a matter of overcoming bias left from the past.

*If you would like to know more about my methodology, email me at a_brewster3@yahoo.com

Posted in Uncategorized | Tagged: , , , | 16 Comments »

Mountain West Expansion

Posted by ncaafbsfootball on March 8, 2009

I always laugh to myself when someone says that if schools like BYU and Utah don’t like being in a conference without an auto-bid they should just join the Pac 10, as if it were just that easy. My personal favorite, however, is when some fans produce some Frankenstein patch-job of a conference with what they perceive to be the country’s top non-BCS teams. Let’s be realistic–the Mountain West will not dissolve, and its bylaws prevent any schools from being kicked out without a unanimous vote to do so. So sorry, but it looks like San Diego St is here to stay. Not to worry, though; the Big XII has Baylor, the ACC has Duke, and so on.

Anyway, the idea of Mountain West expansion isn’t out the window, since the league commissioner stated in 2005 when TCU was added, that the league might be interested in adding one or two schools.  That leaves room for one more.  I’d like to see a move that strengthens the conference and helps secure an auto BCS bid.  I think adding Boise St or Fresno St would be the top two choices for that to happen.  Once the MWC has that bid, I’d like to see the conference strengthened by adding two more teams and a championship game.  My top two choices would be the second team of either FSU or BSU, then if we could steal Colorado from the Big XII, that would really help.  There’s no question that Boise St and Fresno St would jump at the chance to play in the Mountain West, even as is.  I think Colorado might be interested if the MWC were a BCS conference–travel costs would likely be lower, and Colorado would have an easier road at getting to a BCS game through the Mountain West than the Big XII.

Here are some of the other teams frequently suggested, along with my opinion.

Tulsa – This would probably be the best alternative after FSU, BSU, and CU.  Tulsa has shown some decent football teams over the years, but that success has not come against the BCS.  The Golden Hurricane has not beaten a BCS school in the regular season this decade.  Then again, frequent games against OU and Texas creates a difficult road.  Tulsa could serve as a travel partner for TCU.

Hawaii – UH has a decent athletics program, but travel is out of the question.  There is a reason the former WAC schools broke long-established ties with Hawaii when the Warriors were not invited to join the Mountain West.

Nevada -Decent football and basketball programs would bring a great in-state rival for UNLV.  Nevada really hasn’t made any kind of splash on the national scene, though, and would be perceived as making the conference weaker.

SMU – Great travel partner for TCU.  Other than that, SMU’s program is in shambles and bringing in the Mustangs would just create another mouth to feed.

Houston – Not as good a geographical fit as Nevada, and not as good a program as Tulsa.  UH would definitely not be a top choice for the Mountain West.

Posted in Uncategorized | 1 Comment »

The Uphill Battle of non-BCS Schools

Posted by ncaafbsfootball on March 6, 2009

All non-BCS schools would love to have their hand in the BCS cookie jar, but few appear actively interested in developing a football program on par with the average BCS team.  Those reaching for the highest level face a difficult uphill battle.  The current structure of college football forces non-BCS programs to win on the road against quality BCS teams.  This task is statistically insurmountable over the long run considering that the average FBS home team will win over 63% of the time, and virtually no non-BCS conference has a 50/50 home-away relationship with the BCS schools.

In plain English, that relationship means that non-BCS teams and conferences must prove on the field that they are superior to the average BCS team in order to earn esteem equal to the BCS average.  This is the epitome of an uphill battle.

Posted in Uncategorized | Leave a Comment »

The Role of Bias in College Football

Posted by ncaafbsfootball on March 6, 2009

College football’s FBS division now groups 120 of America’s largest programs. With so many teams, head to head matchups of all schools are impossible, which creates the need for speculation. The idea that some teams and some conferences are superior to others is unquestionable; however, the margin of this superiority is entirely debatable.

Currently college football is dominated by the six BCS conferences, which are positioned to leverage fan and media bias to maintain the advantageous status quo. Whenever a team outside the big six challenges that status quo, unfavorable bias is ultimately what bars them from the opportunity and reward they’ve earned. While speculation is absolutely necessary, special attention should be given to accommodate legitimate challengers regardless of conference affiliations.

Posted in Uncategorized | Tagged: , , , | Leave a Comment »