Game of Cannon By Ashish Gupta and Sanket Gandhi
The game description can be found here
The bot part uses Adversarial Search with Alpha-Beta pruning. The search was cut-off according to state analyzer function which decides depth of search for given state. A evaluation function was used to compare between different cut-off leaves.
For choosing next child node to explore we used same evaluation function so that maximum nodes are pruned. The state analyzer function was desgined in such way that at start of game the depth was high and as game progresses the depth decreases.
Eval(state) = 100so + 1000to + 10ct + at - 0.000001dt
where
so = net soldier differencd
to = net townhall differencd
ct = net cannon difference
at = net attacking positions difference
dt = net difference of sum of distances of soldier from opponents townhall
