Actually this doesn’t surprise me at all. I wouldn’t say it reveals a “flaw” in seeding — what it shows is that there is more separation in quality between teams at the top than in the middle seeds.
Consider the “usual” first 2 games for the 9, 10, and 11 seeds:
9 seed - 8/1
10 seed - 7/2
11 seed - 6/3.
In all cases the average opponent seed is the same: 4.5. But the average quality is not.
Let’s look at an example assuming perfect seeding for the 2022 season, today. We can think of every seed based on the KenPom average efficiency margin rating of the four teams on that seed line. So, “seed 1” is the average of KenPom ranked teams 1-4, “seed 2” is 5-8, and so on.
By this metric, here’s the strength of the opponent seedlines in question:
1 seeds: +29.0
2 seeds: +25.2
3 seeds: +24.6
6 seeds: +18.0
7 seeds: +16.2
8 seeds: +15.8
This means beating a 1 and an 8 is harder (average of +22.4) than a 7 and 2 (+20.7) or a 6 and 3 (+21.3).
Of course, there’s also some difference in the 9/10/11 teams themselves:
9 seeds: +15.4
10 seeds: +14.9
11 seeds: +14.4
This clearly doesn’t outweigh the jump in opponent strength going from 7/2 to 8/1.
Add to this that 1 seeds almost never lose in round 1, while 2/3 seeds somewhat commonly lose, and you can see why the 9 seeds usually leave earlier than 10s and 11s. Moving up to the 9 line gets you a slightly easier first game in exchange for a much harder second game.
Of course, in any given season, there’s a lot of extra noise, and it could certainly be that the committee’s system tends to bias the real seeding away from this. The point is that even with ideal seeding—perfect ordering of the teams based on actual quality—it can be easier to make the Sweet Sixteen as a 10 or 11 seed, because the quality of teams does not change linearly with ranking.
