Author: Robert Hyatt
Date: 07:19:10 08/07/99
Go up one level in this thread
On August 07, 1999 at 01:35:31, Bruce Moreland wrote: > >On August 06, 1999 at 20:01:16, Robert Hyatt wrote: > >>On August 06, 1999 at 17:56:35, KarinsDad wrote: >> >>>On August 06, 1999 at 13:59:10, Robert Hyatt wrote: >>> >>>[snip] >>>> >>>>It is a _long_ way from easy. The path has to go at least 100 plies beyond the >>>>first capture/pawn push to make sure you won't run afoul of the 50 move rule >>>>down that path. And to search that path is just a perfectly normal search, >>>>which translates into _impossible_ to do... The databases don't contain trees, >>>>they just contain a score for each possible position that can be reached... >>>> >>>>and I don't see any way to fold in the 50-move rule because of this... hence the >>>>other thread and DTC scores... And even that won't work if the computer has to >>>>'take over' a position played by a human, because then you can't assume optimal >>>>play since reaching a database win, and at that point, the 50 move rule is >>>>death to any sort of probe... >>>> >>> >>>I now understand the difficulties involved (the "take over" case is probably not >>>even worth discussing). >>> >>>My recommendation: punt. >> >>I agree... but note that many are not going to like this, perhaps yourself >>included, when you want to analyze some endgame you played to see what will >>happen after the game is resumed... But I don't see a real solution to this >>problem... >> >> >> >>> >>>Chances are, even if you are playing another program with TBs, you will rarely >>>find move by rep situations or violations of the 50 move rule once you get to >>>the TBs. So, if you get to the TBs, it will either be a lose, a draw, or a win >>>by default and the chances of it changing from a win to a draw (or a loss to a >>>draw) due to either of these two rules should be extremely slim. >> >>that's debatable... since the losing side also has TBs, it will certainly >>put off the loss as long as possible, which might cause problems (ie in Eugene's >>kbn vs kn mate in 100+ position). And if the DTC value is ever added, I'd >>certainly be checking it when I was losing, to always take a move with DTC+ >>current-50-move-counter > 50, to ensure a draw in a technically won ending. :) >> >> >> >> >> >>> >>>Note: The move by rep problem can be minimized (but not eliminated) by using a >>>little AI on choice of moves out of the tablebase. For example, if the white >>>king has only been on f4 since the last capture or push (the reset for 50 move >>>and for move by rep), then you could search the TB for moves that still mate, >>>but start moving the white king away from f4 (or somesuch). >>> >>>Do you have statistics for Crafty on percentage of times (against computer >>>opponents with TBs only) that a win (or a loss) was changed to a draw due to >>>this (either for Crafty or for it's opponent)? >>> >>>KarinsDad :) >> >> >> >> >>I don't believe it has ever happened, at least not that I have seen. > >That KP vs KNN mate in 106 or whatever may have been a case of this. > >If you guys are talking about how to avoid 50-move rule in a DTM database, the >most obvious way to do this is to use a DTC database (not Thompson's, which >doesn't count a pawn move as a conversion, I believe) to turn every won position >with a DTC of more than 50 into a drawn position in the DTM database. > >bruce That would work, but someone tossed in a real clinker... you have to play this ending and win, even if a human was playing the game and had played several sub-optimal moves in the 5 piece ending _before_ the program gets to see the position. Since non-optimal moves have been played, the DTC database will not fix the file properly, because it won't realize that this is a (pawn push conversion) position that requires a pawn push "earlier" than normal due to the extra moves already played since the last pawn push... The deepest mate I have seen so far was mate in 103, but within 5 moves it was under 50, because the opponent didn't make the best moves. However, Eugene's KBN vs KN is even worse (no pawns)... that one is hopeless...
This page took 0 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.