GameKnot: Chess related, chess engines on endgame play

Chess related: chess engines on endgame play

« Back to forum

From

Message

wschmidt
11-Jan-11, 09:26

chess engines on endgame play

I know that chess engines have a poor reputation generally for strategic endgame play (as opposed to simple technical wins). But I'm wondering if anyone knows whether any of the commercially available engines have a better reputation for that area of the game than the rest. I'm looking to replace my ol' Fritz 8 and strong endgame play would be a plus factor in the next engine I get.

Any thoughts? Thanks in advance.

chess_avenger
11-Jan-11, 12:19

just a note:
these are not allowed on gameknot during a game (at least i dont use it).

you can get a nalimov or any other endgame database with 3,4,5,6+ pieces out there.
if you want calculating power rybka or stockfish are probably the best engines (last i checked) which perform okay endgames without any endgame databases.

wschmidt
12-Jan-11, 09:58

Thanks, CA,

I'm not looking for analysis of on-going games and I use the tablebases on-line when I'm doing post-game analysis if things are down to just a few pieces. I was more wondering if any of the engines have a better reputation for endgame analysis when there are lots of pieces and pawns still left on the board.

ganstaman
12-Jan-11, 12:39

I don't know which engine, but I do know that the engine should have the tablebase installed or else it will make some pretty wrong evaluations.

fmgaijin
12-Jan-11, 16:26

Even with Tablebases . . .

. . . most programs evaluate poorly in the ending, particularly R + P endings or Q + P endings. They don't "know" things like "wrong color RP for B," "R behind opponent's extra passed P," "R + RP & BP vs. R draw," "Blockade draw," and so forth UNTIL IT GETS INSIDE THE MOVE HORIZON. So it will blithely go about a R + P ending evaluating it as "win" or "big advantage" when a human master knows that the position is a long-term draw. Then watch the evaluation suddenly drop a few moves before the tablebase kicks in *grin*!!

Likewise, they are weak at endings where one must repeat a pattern over and over again to advance (e.g., advancing a pair of connected passed pawns in an opposite color B ending. It may repeat moves instead of following the pattern if the pattern is too deep.

And, yes, both chess engines and tablebases are not permissible on GK, though I am sure that some players use them . . . and sometimes I wish that I did (made two bad endgame mistakes playing for the GK team "Peace: The Power That Preserves" in the ICCF Champions League, one of which cost me half a point already (W to D) and the other will probably do the same (W to D). I went back and checked the first one with Rybka using tablebases afterwards and found that (even though it evaluated the position poorly), it wouldn't have made the mistake I did (*sigh*). I haven't checked the other one yet (still trying to win despite my mistake), so I'll let you know how Rybka does with that one.

« Back to forum

chess online