carbocarde / battleship Goto Github PK

Hi,
Thanks for nice blog write up on problem I was trying to solve some months ago.
As for solver:

What solver tries to achieve is to create heat map of probability that part of the ship occupies the spot. There are two approaches here:

• iterate over each of ships, try to place them on the board, count times spot is occupied by ship. This is current approach in solver
• iterate all permutations of available ships and create heat map - how ships can be placed on board is function of non-occupied places, number of ships and rules which gives much more accurate heat map,using tremendously more computation than former approach.

Lets take a example:
3x3 board, available ships 1x3size + 2x1size, board initial conditions:

0 0 0
0 0 0
0 0 0

• with current solver, quite uniform distribution (counting 1x3size ship + 2x1size ship)

4 4 4
4 4 4
4 4 4

• with alternative: 3size ship can be place only on one side (ex horizontal bottom) with 2x1 size on only top-right and top-left corner

4 1 4
1 0 1
4 1 4

From probability side, improved heat-map is more precise when we reduce available space in board (either by sinking ships or by miss shots) ,also as it is heavy in calculating permutations, should be done on reduced ship size and board size - thus during mid-to-end game.

I was not able to create performant solution of such solver, I will leave that challenge to you ;)
Keep it up!

Description

The following behavior has been seen when toggling squares between hit or miss:

• The circle for suggested next move does not change position.
• The heatmap for probability of each square does not change.