You should start mentioning overall training levels at the time of the watched/unwatched games, I won an association game of 22% Q discrepancy when I had 4 10% levels and have been consistently beating higher Q teams both watching/not watching with training levels of 32%(4x8%) and above. It might give more insight if you do a cross analysis. Unfortunately, not knowing your opponent's training levels will render your experiment fruitless as I believe that that comparison would be more indicative of who would win vs a 10% possession swing.