NCAA March Madness Analysis: Wins Over Bubble

The NCAA Tournament begins a month from tomorrow. The bubble talk has already begun, and the RPI and Joe Lunardi will be ever-present over the next thirty days. Last year, I experimented with something I ended up calling “Wins Over Median At-Large”.

That name, though, was clunky, and had a point of reference of roughly the 25th best team in the country. After reflection, I changed the point of focus to the bubble–the cut line for where a team should be considered for at-large entry. In the NCAA tournament, the play-in at-large teams generally fall around the 12 seed line. Thus, I am using the 50th best team in college basketball as my reference.

So now, we have “Wins over Bubble” (WOB). The goal with WOB is to cut through all the nonsense with the RPI and provide an alternate way of assessing accomplishment. The RPI way over-rewards playing decent but still extremely beatable opponents at the expense of playing really poor teams. It does not do a good enough job, at the other extreme, of differentiating between playing a top 5 team on the road, and the 24th best team at home. WOB seeks to answer the question, “How many wins would a true bubble team be expected to have against this schedule?”

To answer this, I utilized the Pomeroy rankings, differentiated home/away/neutral, and calculated an expected win percentage based on opponent strength and venue (rounded to nearest 5%) for each game. Let’s give some examples of why this is important. You could play Fairfield (#283 in Pomeroy, and 5-22) and be expected to win 95%, or play Florida Gulf Coast (#183, 17-10) and be expected to win 90% of the time. Because of their different records, just swapping that one game will have a big impact on RPI.

So these rankings reflect the expected record against a team’s schedule for a true bubble team with average luck. They do not reflect point differential, so like the RPI, are supposed to be retrodictive and not predictive. It’s great for evaluating schedule. Kansas is the number 1 schedule by RPI. They have played a difficult schedule, though Michigan (4th according to RPI) comes in slightly harder by this method. A true bubble team that played Michigan’s schedule to date would be expected to be 13-12. (A bubble team against Kansas’ schedule would be expected to win 13.2 games out of 25). Michigan going 18-7 against that schedule is as impressive as Cincinnati going 23-3.

Here are the results. The current cut line for at-large selections would fall between #52 (St. John’s) and #53 (NC State). Mobile Link here