paulsonkoly/chess

A UCI engine written in Haskell

chess's Issues

The engine sometimes dies when there is an unavoidable mate

I've seen this happening couple of times. Needs debug.

threefold repetition / 50 move draw are not implemented

threefold repetition : http://en.wikipedia.org/wiki/Threefold_repetition
50 move draw : http://en.wikipedia.org/wiki/Fifty-move_rule

Both of these can be implemented using a history of the Zobrist hash. Question is where that history should live - ie part of the board or some other place that stores game history.

QuickCheck

I collect here the different ideas of what could be quickChecked.

• magic lookup gives the same as slidingAttackBB
• making a move on a Board with n pieces results in a Board with n or n - 1 pieces
• recalculating Zobrist from scratch is the same that the move modified hash

Shorter mate is better than longer

We don't order the mating sequences. Therefore it is possible that depending on the order at which they are discovered we choose a longer sequence. This results in pointless in between moves that don't help the mate

Also as we don't have the 3 - repetition rule implemented we might draw when we see a mate.

Need time control

On better CPUs NegaScout easily gets down to depth 6-7, on old machines it might struggle with 5. We need a module that drives the depth based on the game time control.