Source: http://fivethirtyeight.com/datalab/do-co...

The admitted flaw in both approaches is the same one I ran into when evaluating which college basketball coaches outperform NCAA tournament expectations based on seeding. Such a method ostensibly captures underperformance in the final game of a tournament, but it doesn’t detect the missing future wins expected of a favored team going forward. Moreover, these results shouldn’t be taken to say that conference tournaments have no predictive value. A team’s post-conference-tourney SRS is slightly more correlated with its eventual NCAA tournament wins than its rating before the conference tourney began.

However, this analysis does serve as a warning against putting too much emphasis on the conference tournament relative to a team’s entire body of work, especially when it comes to picking unexpectedly hot conference tournament teams to go further than you’d otherwise predict for teams with their résumés.