panchishin/alphabeta

Javascript implementation of minimax in Node using asynchronous calls to customizable game logic, scoring, and move generation.

alphabeta's Issues

How to handle loss and draw conditions with only checkWinConditions?

I am working on a game that ends when all squares on a board have been filled. Each player much fill a square whenever it is his turn.
The player wil the most squares of his color is the winner. Equal amounts of occupied squares indicate a draw. How should I handle this with this package?

Generating next possible states is doing to much work

As a rule of thumb alphabeta roughly has a branching factor of the sqrt(number of possible moves). So because this implementation generates all possible states it does a lot of unneccessary work. It would be better to generate moves and then when needed perform/play the move and generate the new state.

It can them be decided if undoing/takingback the move on a single state is faster/slower then generating new states.

Introducing moves has the other great advantage that it is easier to sort them in an order that speeds of the alphabeta algorithm. Keeping a best move per depth (killer move heuristic) and trying that move first for instance can greatly reduce the searchtime.

implement killer move

This function could be really useful but as an absolute minimum it should try the best move from a depth search first during the next (deeper) depth search. Because in that case there is always a good move available when result.incomplete is returned.

It would be very good for the algorithm speed to keep a last best move for each depth. (see also: #2)

BTW "incriment" should be spelled "increment"