March Madness: Does Conference Tournament Timing Matter?

By Brendan Kent

This yearâ€™s Big Ten tournament was a week earlier than usual, ending a week before Selection Sunday rather than on Selection Sunday. The conference wanted the tournament in Madison Square Garden, but thatâ€™s Big East Tournament territory, so the Big Ten had to trade timing for setting.

Some have wondered whether this move could disadvantage Big Ten teams heading into the NCAA Tournament:

So, should you be wary of teams that played an early conference tournament when making your First Round bracket picks?

To find out, I dug into Kaggleâ€™s excellent March Madness data set, which contains NCAA Tournament, conference tournament, and regular season games from 1985 to 2016. I teased out the number of games between a teamâ€™s final non-NCAA Tournament game (either a conference tournament game or a final regular season game, if the teamâ€™s conference did not hold a tournament) and its first NCAA Tournament game, essentially a teamâ€™s pre-NCAA Tournament rest days.

I then tested whether the number of rest days had any impact on a teamâ€™s performance in its first tournament game (either a First Round game or First Four game) by regressing an indicator for whether a team won its first game on difference between the opponentâ€™s seed and the teamâ€™s seed, the year, and the team’s rest days. Iâ€™m considering First Four games in this analysis because those are the first NCAA Tournament games for the participating teams, even though the games arenâ€™t First Round games and really shouldnâ€™t even exist in the first place, but Iâ€™m getting off topic.

In addition to testing for a simple linear relationship between rest days and winning the first tournament game, I considered a quadratic relationship. In theory, if there were an optimal number of rest days that was neither too few nor too many, the relationship might be evident via a quadratic fit.

The regressions (summarized below) found no relationship, linear or quadratic, between rest days and winning a first tournament game, when controlling for seed differential.

But what if rest days impacts some seeds and not others? To check this, I created indicator variables for team seeds and interacted them with rest days, again treating rest days first a linear variable and then as a quadratic variable. I also removed all First Four games and First Round games for teams that played in the First Four for this test because in the First Four, teams play opponents that share their respective seed, so the indicator for seed does not imply the seed of the opponent, as it does in the First Round (i.e. a one-seed always plays a 16-seed, etc.). I wonâ€™t bore you with the full (very large) regression table this time but, in short, there is still no evidence of a relationship between winning a First Round game and rest days when you model seed this way.

This is not to say that conference tournament timing has no effect on teams. It may just be that teams are able to adequately deal with having either too much or too little rest between conference tournaments and the NCAA Tournament. Although unlikely, it could also be that unusually long or short breaks impact a teamâ€™s performance later in the tournament, and not in the first game. Regardless, itâ€™s probably safe to send your Big Ten picks through to the second round this year without concern for their unusual conference tournament timing.