College Football's 2020 "Bad Data" Problem

Updated February 21, 2021

The coronavirus pandemic significantly altered team schedules in advance of and throughout the 2020 college football season. Conferences established competition safety protocols and scrapped many pre-existing schedule arrangements.  Several conferences opted to delay and shorten the season, while others filled the fall calendar with more conference games than ever before. Teams announced new game matchups merely weeks (and in some cases, merely days) before kickoff. Three FBS programs opted out of playing altogether.

Virus case counts and contact tracing procedures within programs forced the postponement and/or cancellation of numerous games each week. Some teams navigated the year with few disruptions, playing as many as 13 games. Others were less fortunate, playing as few as three. In total, the 2020 season featured 534 games between FBS opponents, 240 fewer games than were played the year before.

The drop-off in non-conference game scheduling was concentrated in some conferences more than others. There were only 19 games played between Power 5 (P5) and Non-Power 5 (NP5) opponents. There were only 12 non-conference games played between P5 opponents, none in the regular season. Instead of an abundance of connecting threads binding FBS teams together, the 2020 season featured predominantly isolated islands of conference competition, sparsely linked by a limted number of non-conference games.

College football's 2020 "bad data problem" is somewhat attributable to the strange circumstances of the games themselves (COVID absences, crowdless stadiums, etc.) but I suspect much more attributable to an extremely disconnected network of teams, and the impact of that on effective opponent adjustments. I do not think 2020 FEI data is completely worthless, but it is an unreliable outlier in comparison to data from other seasons.