Als Antwort auf:/In reply to: Re: I will bite........what of it? geschrieben von:/posted by: Dieter Bürßner at 17 May 2004 21:19:13:
Regarding KRKN in Bestia see my old messageRichard, I agree in general. But KQKN is really easy in this context. Just eval it as win for the Q, and wait until the pos is on the board, when TBs take over. It would not matter, if the eval accounts for a really good KQKN position or a difficult one, when you decide to trust the TBs later (many other positions are similar). Of course you argument is important, when the position/material constellation is not that clear.Of course, when the game gets down to 4 pieces, the EGTB's would take care of it.
But ...
Probing the EGTB's in search is costly. Disk access is slow. That is why most (if not all) programs do not probe in the quiescense search. And guess what: The quiescense search is all about evaluating what happens after a capture sequence. Also what happens after a capture sequence that leads to an EGTB position. Probing the EGTB's are too expensive, so the evaluation should know if a position is favourable or not.
So endgame knowledge of basic endings is still important. It may find key moves in an ending a few plies earlier.
For aesthetical reasons, I would prefer my engine to be able to win KQKN without TBs. I did not try yet, but will soon ... If it doesn't, I will probably fix it, even when it is no issue for playing strength under my typical setup.
This reminds me of KRKN and KRKB. I remember Gearge posted here once some interesting ideas about it. Does anybody find those posts? For these positions, better knowledge could also help, even when using TBs.
Regards,
Dieter
With KBKR Bestia uses the same table as for KBNK to avoid king moving to the bad corner + tries to keep distance between kings.
Seems to work fine...
Best regards,
George