Shame

[Henk]

I just found out that my engine can not finish a KQK end game. PSQ table says that King should stay in center. Don't know if that's the only reason. My engine also does not know that KK is a draw.

[lauriet]

Hey, you better fix it

[Henk]

Hey, you better fix it

Don't know if there is an easy way. Less code the better. Also it should not slow down search in normal positions.

[Bo Persson]

I just found out that my engine can not finish a KQK end game. PSQ table says that King should stay in center. Don't know if that's the only reason.

At least it's one reason. Your Queen needs the support of her king to be able to mate the other king along the edge of the board. They should both close in on him to restrict his movements.

My engine also does not know that KK is a draw.

So there is some room for improvement.

When fixing this, you might consider that without a Q or R, you probably have a draw long before reaching KK.

[Henk]

Strange alpha beta search with TTable is able to solve this easily while PVS can't. Difference is only three lines of code. After a few moves the algorithm should see a mate in eight or so.

[d] 4k3/8/8/8/8/8/8/6KQ w - - 0 1

[Henk]

Even simple alpha beta search can go wrong. I forgot to try best move from previous iteration first but returning best move found so far that is stored in TTable. So once in a while if time is up you get really bad moves. For search did not end and it did not try move from previous iteration first.

Also mate scores may not be stored and retrieved right from TTable.

[Sven]

Strange alpha beta search with TTable is able to solve this easily while PVS can't. Difference is only three lines of code. After a few moves the algorithm should see a mate in eight or so.

Let us guess more of these three lines, then we can try to guess the bug ...

[Henk]

So perhaps after all possible perft tests has been done. Next step is to solve mate in N problems with alpha beta search and no TTable.

Success is a choice.

[Aleks Peshkov]

Perft with TT helped me to detect bugs that would be very difficult to discover otherwise.

[Henk]

Strange alpha beta search with TTable is able to solve this easily while PVS can't. Difference is only three lines of code. After a few moves the algorithm should see a mate in eight or so.

Let us guess more of these three lines, then we can try to guess the bug ...

My guess is it that it had something to do with TTable. As always. No TTable no bugs.