UConn/Butler: Is the AP Poll Losing Predictability?

This year’s NCAA tournament has been one of the most unpredictable in recent memory. VCU made it to the Final Four, and now UCONN and Butler are playing in the national title game after UCONN was unranked in the pre-season poll and Butler was unranked for all but the first two weeks of the season. This raises the question: how accurate are the AP polls, particularly the pre-season poll? The table above shows how many of the teams in the pre-season Top 25 ended up being in the final Top 25, since the 1989-1990 season. This way of measuring the accuracy of polls is probably the most favorable to the pollsters. It does not take into account shifts within the top 25; that approach would certainly judged the pollsters in a more harsh manner.
The pre-season polls reached the peak of their accuracy in during the 1999-2001 period and has reached its low during the current three-year period 2009-2011. The trend in polling accuracy was relatively consistent until around 2007, the same year as the start of the one-and-done rule. Before that rule, players that attended college usually stayed more than one year, allowing the pollsters to gain more familiarity and evaluate teams accordingly. Connecticut was unranked in this year‘s preseason poll, no doubt because it had a number of unknown players; pollsters did not know how to evaluate them.