The Power of the Preseason Poll

With a relatively short season, a single elimination tournament, and 18-year old athletes, college basketball is one of the hardest sports to measure and predict. Sources like Ken Pomeroy and ESPN’s BPI have been huge improvements in recent years, but the flaws of the RPI and the AP Poll are still extremely visible and accepted far too easily.

For this piece, I wanted to study the AP poll and see just how effective it was at predicting tournament success. In recent years this has been especially strong, as Kentucky this year (#1 preseason, unranked postseason) made a run to the final 4 and knocked off Wichita St. (#16 preseason, #2 postseason) in the second round. So just how strong is the correlation between the first poll of the year, the last poll of the year, and the tournament to end the year?

A lot of the voters are old-timers with poor reasoning and shoddy analysis, and it turns out that the pre-season poll is nearly as effective at predicting March Madness games as the end of season poll is. I went back the last 15 years and tracked the preseason and postseason polls and compared them to tournament wins. Here is the raw poll data:

Note: if a team that went far in the tournament is missing, it’s because they weren’t in the top 25 of the first or last poll of the year.

Some of the more interesting trends I found:

• Of the last 14 champions, 3 have been the preseason #1 team and 2 have been the postseason #1 team.
• Of the last 14 champions, 3 weren’t ranked in the preseason (2003 Syracuse, 2006 Florida, & 2011 Connecticut).
• Of the four teams to start AND end the year at #1 in the poll (2002 Duke, 2006 Duke, 2008 UNC, & 2010 Kansas), only one made it past the Sweet 16.
• The preseason #1 teams average 3.13 tournament wins. None have missed the tournament, and only 2005 Kansas lost in the first round.
• The postseason #1 teams average (you guessed it) 3.13 tournament wins. All of them have won their first round game.
• Preseason top 5 teams average 2.73 tournament teams, while postseason top 5 teams average 3.01 tournament wins.

When you account for the fact that postseason ranked teams have higher seeds and easier games, it’s pretty impressive that the preseason ranked teams are able to come close to matching their win totals. A major flaw in the polls is the lack of respect for strength of schedule and rewarding non-conference teams that run through their schedules. By the end of the year, the likes of Wichita St. and Gonzaga often shoot up farther than they should while the 3^rd and 4^th best teams in power conferences are downgraded too much. In recent years this has been even more apparent with the raw talent of teams like Kentucky getting under-seeded and teams like Saint Louis riding an easier schedule and worse underlying numbers to a high seed.

As a whole, the preseason data had a .402 correlation to march madness wins, and postseason data had a .503 correlation coefficient. Below is a chart of how the two have compared from year to year.

Looking into next season and whenever evaluating college basketball, remember to look at the right metrics and know what to value. And going into next year, don’t downplay the power of the preseason poll.