Here's how the prelims would work.
- Let all teams with a winning record (four or more wins in seven games) qualify for the playoffs.
- In rounds 1-5, teams play other teams with equal records. By the binomial theorem, this gives one 5-0 team, one 0-5 team, five each at 4-1 and 1-4, and ten each at 3-2 and 2-3.
- In round 6, pair 3-2 and 2-3 teams against each other. This can be done without repeating matchups and gives five 4-2, five 2-4, and ten 3-3 teams.
- In round 7, pair 3-3 teams against each other, giving five 4-3 and five 3-4 teams. This gives a total of 16 teams left in contention after the prelims.
- In rounds 6 and 7, pair any team with at least four wins or at least four losses -- any team already eliminated or qualified -- against a team with the same record in the other division. Play, perhaps, games for four-win teams in round 6 on the Division I packet and those in round 7 on the Division II packet, and vice versa for four-loss teams.
- Break the 16 qualified teams into four pools of 4 and play a bracketed round robin, followed by tiebreaker if needed;
- Advance the top two from each pool together into one of two pools of four and play a bracketed round robin, counting the game from the previous phase between the two teams, followed by tiebreaker if needed;
- Advance the top two from each pool together into one pool of four and play a bracketed round robin, counting the game from the previous phase between the two teams, followed by tiebreaker if needed;
- Play a NAQT-style final between the top two teams.
Is there a reason we couldn't pull off a scheme like this, or that it would be inadvisable? Would it be valuable enough to have some actual data on how Division I teams performs on Division II packets and vice versa (say, for calibrating the D-Value conversion factor) that it would be worth trying?