I spend a lot of time on sim-ing for possible temporary Filter structurs on the top 3 rows for optimal attacks.
And a lot of effort in sorting them by potential … caching good solutions etc. (I was limited to 200-500 sims)
1 in 5 moves was cool, 1-20 moves was magical … but more often then not, random enemy actions can just trash a perfect attack plan.
After one month of working on this, my iniatial algo that was just stacking and sending normal attacks was with better ranking
So just sim-ing for the perfect build is not enough. You need to take in to account for possible builds of the enemy, history of attacks, how good he reacts to changes … and probably a lot more.