Author: Robert Hyatt
Date: 12:24:26 10/09/98
Go up one level in this thread
On October 09, 1998 at 15:17:43, blass uri wrote: > >On October 09, 1998 at 14:56:46, José de Jesús García Ruvalcaba wrote: > >>On October 09, 1998 at 13:04:53, John Coffey wrote: >> >>>On October 09, 1998 at 09:38:35, Robert Hyatt wrote: >>> >>>>On October 09, 1998 at 02:11:19, Jouni Uski wrote: >>>> >>>>>I have feeling, that using 3-5 pieces tablebases don't give any measurable >>>>>rating gain - may be 2-10 points maximum = no real gain. >>>>> >>>>>Jouni >>>> >>>> >>>>I would disagree here, as I see KRP vs KR regularly. In fact, in long >>>>games on ICC, I'd bet I see this about every 10 games or so. Lonnie >>>>can comment because he has played against Crafty a lot. And it is quite >>>>often for two computers to end up a pawn up or down, in the ending, and >>>>if one knows about this, it will win more, or draw when it should lose, >>>>than without them. >>>> >>>>ditto for endings like KRB vs KR... where it knows to trade into that >>>>when it can be drawn... >>> >>>Fritz5 has KRP vs KR and KQ vs KR. Not sure if it has anything else. >>> >>>I wanted to create a tablebase with all the possible K + 2 pawns (or less) >>>vs all the K + pawn positions. I figured that the number of possible positions >>>is around 489 million, but if I were to store it 2 bits per position >>>(win/loss/draw/unknown) then I could get it in 128 megs. >>> >>>John Coffey >> >>To make a tablebase, you need all of its descendants first. The direct sons of >>KPPKP are: >>KPKP (black captures one white pawn) >>KPPK (white captures the black pawn) >>KQPKP (white promotes to a queen) >>KRPKP (white promotes to a rook) >>KBPKP (white promotes to a bishop) >>KNPKP (white promotes to a knight) >>KPPKQ (black promotes to a queen) >>KPPKR (black promotes to a rook) >>KPPKB (black promotes to a bishop) >>KPPKN (black promotes to a knight) >> And each one of these has its own sons, which must be calculated first, until >>you reach KK, KBK or KNK (which are drawn). So, your storage requirements might >>be a little higher. >> With so many pawns, a tablebase with little information (like the one you want >>to make), could be useful, as you can search only the branches which do not >>decrease the exact score for the side to move. > >I think that practically you can do some assumptions to decrease the number of >endings. >for example KQPKP if white to move and the black only P is not close to be a >queen is a win for white. > >white promotes only to a queen and not to a rook,bishop or knight >practically the cases when white is winning this endgame by underpromotion are >very rare. > >Uri This doesn't matter however, because *every* possible position must be accounted for, with an exact distance to mate for the side on move with that specific piece configuration. So they *all* have to be computed to build the next one after them...
This page took 0.01 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.