Results: https://docs.google.com/spreadsheets/d/1KjJ3bMQH66ODTKRkIpqdkWQmFD9JqWUk8Hj8VHM_ftA/edit?usp=sharing
Out of 14 trials, ABBABAAB was better 5 times, ABBAABBA was better 9 times.
Method:
Used PST-Sun sign-up player pool. I believe that gives a typical mmr distribution. For each trial, randomized mmr order, and used the first 8 entries as a draft pool.
Conclusion and comments:
ABBAABBA is better than ABBABAAB in typical cases. These schoolyard formats are not good at making balanced teams. Real matchmaking, with ratings that update, is preferred. If “hidden mmr” is accurate (no smurfs or anti-value players), balanced shuffle should produce much fairer teams than schoolyard drafts.
Public Notice 