I am trying to solve the following game http://www.hacker.org/brick/ using some kind of AI. The target of this game is to finally clear the board by clicking on groups of at least 3 bricks of the same color and removing them. If a group is disappearing the remaining bricks above will fall down or be moved left if a column got no bricks left. The higher the level -> more colors and larger board.

I already guessed that a pure bruteforce approach wont scale nice for higher levels. So i tried to implement a monte carlo like approach which worked ok for the first levels. But i am still not confident i will make the maximum level of 1052 with this. Currently i am stuck @~ level 100 :) The finding of the solution takes too much time...

Hoping that there is a better way to do this i read some stuff about neural networks but i am really at the beginning of this. Before becoming obsessed by ANNs i want to be sure it is the right way for my problem.

So my question is: Does it make any sense to apply an ANN to this game? Any suggestions?