NCAA Tournament: What Happens When the Selection Committee Seeding Differs Greatly From Pomeroy Ratings?

Last year, the Connecticut Huskies finished 8th overall in Ken Pomeroy’s ratings after their run to an improbable national title. That’s not, however, where they were entering the tournament. They sat at 25th, very similar to where the Selection Committee seeded them as a 7-seed.

Kentucky, along with fellow conference member Tennessee, meanwhile, were very much a team rated higher by Pomeroy’s ratings than where the Committee placed them. To much controversy, the Committee put Kentucky as an 8-seed (18th in Pomeroy) and Tennessee as an 11-seed (14th in Pomeroy). Tennessee advanced to the Sweet 16 before losing in close fashion to Michigan; Kentucky went to the title game.

On his site, Pomeroy has now posted for download the pre-tournament rankings, so we can assess how teams were viewed before the NCAA Tournament matchups.

So, of course, now we can assess how teams fare when there is a large chasm between how the team is seeded (based on the Committee’s view of strength of resumé), and how Pomeroy’s system assesses them.

For most teams, the two are similar. There is little disagreement near the top. Every team that has been a #1 seed in the last eight years was in the Top 7 in Pomeroy’s pre-tourney ratings. Only one #2 seed has been outside Pomeroy’s Top 15 (Florida, 2011, 17th).

So in most cases, there is no discord. What happens when there is? Using the pre-tournament rankings, I compared the two. I found all tournament teams and assigned a “Pomeroy seed” based on where they were in the ratings (top 4 teams as 1 seeds, next 4 as 2 seeds, and so forth). I then found those that were at least 3 seed lines apart between the Pomeroy seed and Selection Committee’s actual seeding.

Among the teams seeded as a 12 or better (i.e., where most at-larges fall), there were 45 “Over-seeded” teams (from the perspective that Pomeroy ratings would have seeded them worse), and 50 “Under-seeded” teams in the last eight tournaments. That’s roughly 5 to 6 per tourney of each variety.

The teams were, in aggregate, almost mirror images. The Over-seeded teams were on average a 6 seed, but Pomeroy had them rated more like a 10 seed. The Under-seeded teams were on average a 10 seed, but Pomeroy would have them as a 6 seed.

Using the average win totals by seed in the NCAA Tournament, the Over-seeded teams would have been expected to have 49.1 wins. They had 36 wins. And lest you think I picked 2007 as the cutoff as an endpoint to make the data better, that was the best year for the Over-seeded teams. All seven won in the first round, and three advanced to the Sweet Sixteen.

Since then, only 4 of the 38 Over-seeded teams have advanced to the Sweet 16, even though they ranged in seed from 2 to 9, with an average of a 6 seed. Here’s a summary of how far the Over-seeded teams have advanced, by their tournament seeding.

That #8 seed is Butler, knocking off Pittsburgh, winning some wild games, and advancing all the way to the title game. Absent Brad Stephens’ magic, though, the Over-seeded teams tend to struggle.

Let’s go to the Under-seeded teams. The expected wins are 34.1 based on seeding; they’ve actually won 39. That’s good, though not extreme. Of course, five of those first-game matchups involved games where both were considered under-seeded, so someone was walking away with zero wins. Take the lower-ranked team out of those matchups, and it goes to 30.2 expected and 38 actual wins.

The other issue is that they tend to be overrepresented in losses to top seeds. The under-seeded teams went 2-13 in games against #1 seeds (usually after winning a 8-9 matchup), so over a quarter were eliminated by top seeds.

In head to head matchups in the first game, under-seeded teams are 6-2 against over-seeded teams, while being seeded worse by the Committee.

Here is the performance by seed of the teams who would be considered under-seeded going by the Pomeroy rankings.

Why do 12 seed do so well in the tournament? Well, for one, the Committee’s method of seeding often puts a power team in recent years (like Texas) in a position to make a run out of that spot. Davidson, with Steph Curry, is that 10-seed to the Elite Eight. Kentucky is the Final 4 team.

Some of it is going to happen. Some is because of the wonky way in which we assess “resumé”, over-emphasizing a big win by a team that had a hot night, versus a team showing they can easily lose to sub-tournament teams on a regular basis. Some of it is the vagaries of the RPI and how some teams can game the system to look better than they are by scheduling okay, but not great, not horrible teams. Some of it is close game luck.

Who are candidates this year?

Arkansas (31st) and Maryland (32nd) are both projected as top 5 seeds by many. The reasons there, particularly with Maryland, are simple. The Terps are 10-1 in close games decided by 7 or less. They’ve only lost 5 times, but by an average of 15.4 points, which is almost unheard of for a top team. VCU (36th) could be depending on how the Committee treats them. Seeding them above a 7 seed would be a RPI creation; they haven’t looked great for two months.

At the other end, the likely candidates are Texas if they get in (19th in Pomeroy, 11 or 12 seed), Davidson if seeded 10 or lower (24th), BYU (30th) if seeded as an 11 or 12.