There’s a mistake in this language that we want to clarify. First, a round robin group (or pod) will have 6 teams in it. Second, to create these pods, we need to create “pots” – I’m sure now you see how this mix-up happened.
Each pot has 4 teams who are sorted by rank, 1-4, 5-8, etc. Then they are randomly assigned to a group. That means each group will have one team from each pot, spreading out the seeds among each group with some variance created by the random assignment.
I’d also like to add that we’ve considered a deterministic system that places teams in a “snake” pattern, where 1-4 are placed, then 5-8 in reverse order. That means groups have some pairs of seeds, but it also spreads out the seeds evenly in every group. So, for instance, 1 would be in a group with 1, 8, 9, 16, 17, and 24. 4 is in a group with 4, 5, 12, 13, 20, 21.
I’m curious what the community thinks is a better model. We’re really excited about the chance to use a group stage and we’ve already gotten some excited feedback about it, but we’re still refining, so feel free to make your concerns known.