Uses a random hunt with adjacent targeting. Hunting uses a parity (checkboard) based approach. When a hit is observed, all of the adjacent cells not already visited are searched until there are no more focused targets at which point the hunt continues
60 - 65 shot average per win (not emprical)
Employs a probablistic solution based on density using a little linear algebra. During each round of the game a density matrix is built for the individual pieces still in play by determining all possible positions on the board that a piece can fit. Each element in the matrix is a sum of the # of times the piece could be positioned there in both orientations. All individual matrices are then summed up to provide an overall density picture.
While hunting, the densest positions in the matrix are visited first. If more than one position is found, a random sample is taken from the set. Once a hit is observed the targeting initially uses the density info to order the adjacent cells to attack. After two or more hits, the algorithm will favor vertical and horizontal targets based on the hit pattern. If this can't be determined (no single horizontal or vertical path) the density values are again used to order target positions.
40 - 50 shot average per win (not emprical)
Clone the repo
$ git clone https://github.com/steelThread/battle-group.git
$ cd battle-groupRequires Ruby 1.9, I'm using 1.9.3p286. The player will connect to a local rails server. To change just update the code.
$ ruby battle_group.rb