Reply
 
LinkBack Thread Tools Display Modes
  #1   Report Post  
Old March 31st 04, 08:00 PM
Robert Richter
 
Posts: n/a
Default Endgame tablebase observation--not always harder

I downloaded some endgame tablebases for Arasan, and it is interesting
some of the moves it makes on KRvsK and other simple checkmates. Now
the interesting point is that I never learned the two bishop checkmate
from a textbook--I worked it out on my own and I am likely inefficient
at it. Without the tablebases, Arasan makes me earn my checkmate.
With the tablebases, it hands it to me; my bishops spin somersaults
over each other as I shove the king into the corner. And it is
supposed to "make the best move".

I suppose the reason for this is that if it fought me, there is an
obscure move that I could make that will checkmate it very quickly, so
it chooses not to fight me. I suppose if someone has an inefficient
KRvsK, or even KQvsK checkmate, the computer will likely not make them
work their inefficient pattern and fight them, because it sees that if
it did, there is a way for me to make short work of it, so it doesn't
always seem to want to fight.

I verified this by switching sides and putting up the fight I thought
it ought to do, and it suddenly takes moves I didn't expect and has me
mated in no time.

However, the point is, I was surprised after downloading these
tablebases that Arasan no longer puts up a fight on the two bishop
checkmate.
  #2   Report Post  
Old March 31st 04, 08:12 PM
Mikko Nummelin
 
Posts: n/a
Default Endgame tablebase observation--not always harder

On Wed, 31 Mar 2004, Robert Richter wrote:

However, the point is, I was surprised after downloading these
tablebases that Arasan no longer puts up a fight on the two bishop
checkmate.


Is there really a fight to do in that case? Just force the enemy king to
the edge and then to the corner and that's that. Of course computers with
tablebases think that the best fight is the longest. It seems as if you
have memorized only some predictable patterns following the "longest way"
but not taken care to "branch into easier mates" in case of "less-pedantic"
fighter. I suggest that you practice kbbk and kbnk in a way that practice
matches are against computer without tablebases but, if you fail to give
the mate, you _analyze_ the failed attempt with computer with tablebases.
Should work as time passes ...


Mikko Nummelin
  #3   Report Post  
Old April 1st 04, 01:14 AM
Robert Richter
 
Posts: n/a
Default Endgame tablebase observation--not always harder

(Robert Richter) wrote in message . com...
I downloaded some endgame tablebases for Arasan, and it is interesting
some of the moves it makes on KRvsK and other simple checkmates. Now
the interesting point is that I never learned the two bishop checkmate
from a textbook--I worked it out on my own and I am likely inefficient
at it. Without the tablebases, Arasan makes me earn my checkmate.
With the tablebases, it hands it to me; my bishops spin somersaults
over each other as I shove the king into the corner. And it is
supposed to "make the best move".

I suppose the reason for this is that if it fought me, there is an
obscure move that I could make that will checkmate it very quickly, so
it chooses not to fight me. I suppose if someone has an inefficient
KRvsK, or even KQvsK checkmate, the computer will likely not make them
work their inefficient pattern and fight them, because it sees that if
it did, there is a way for me to make short work of it, so it doesn't
always seem to want to fight.

I verified this by switching sides and putting up the fight I thought
it ought to do, and it suddenly takes moves I didn't expect and has me
mated in no time.

However, the point is, I was surprised after downloading these
tablebases that Arasan no longer puts up a fight on the two bishop
checkmate.



Just to add to this--I learned how to remove the last pawn on the 7th
rank with my king and queen against his pawn and king--as long as it
isn't a,c,f, or h rank. After practicing the moves so that the king
and queen never stumbled over each other and I felt confident, I tried
it against the computer. Before I rightfully earned the pawn, his
king fled the scene and left his pawn unprotected; I guess checkmating
six moves later after rightfully earning the pawn was fewer moves to
mate than fleeing, so it let me get the pawn, and then do an easy KQ
vs K mate.

Anyway, how often does a computer refuse a devastating move because it
senses that you have one and only one way out, and the way out would
put it behind a pawn? Is there a chess "rule of thumb" that states
that if you fight, you may be worse off than fleeing? The database in
KBB vs K seems to indicate this, or else my technique is horridly
inefficient. The main reason for this post was that the database
chose to flee my bishops rather than fight, and that surprised me.

As far as KBB vs K being trivial, it is, but my method is probably
inefficient; what I meant was I didn't have to get the king very
involved in driving him into the corner when doing it against the
database, and normally, the king takes an active role in the driving
process.

I notice that KBN vs K is a lot more difficult with database tables.
I still fit it in 50 moves, unless I blunder and blow a net; however,
a beginner like me probably shouldn't be playing around with this one,
but I think I will keep it in my repertoire.

PS. I hope the experts here are patient with beginners like me.
  #4   Report Post  
Old April 1st 04, 06:51 AM
Guy Macon
 
Posts: n/a
Default Endgame tablebase observation--not always harder


Robert Richter says...

I downloaded some endgame tablebases for Arasan, and it is interesting
some of the moves it makes on KRvsK and other simple checkmates. Now
the interesting point is that I never learned the two bishop checkmate
from a textbook--I worked it out on my own and I am likely inefficient
at it. Without the tablebases, Arasan makes me earn my checkmate.
With the tablebases, it hands it to me; my bishops spin somersaults
over each other as I shove the king into the corner. And it is
supposed to "make the best move".

I suppose the reason for this is that if it fought me, there is an
obscure move that I could make that will checkmate it very quickly, so
it chooses not to fight me. I suppose if someone has an inefficient
KRvsK, or even KQvsK checkmate, the computer will likely not make them
work their inefficient pattern and fight them, because it sees that if
it did, there is a way for me to make short work of it, so it doesn't
always seem to want to fight.


This is because "make the best move" isn't well defined. Most
tablebases are programmed such that, in a losing position, it
makes a perfect attacker spend the most moves possible to win.
It would be far better to make replies that offer the attacker
the most opportunities to make a mistake, even if doing so lets
a perfect attacker win quicker. In chess we don't get extra
points if we make it take someone longer to turn a won position
into a checkmate.


--
Guy Macon, Electronics Engineer & Project Manager for hire.
Remember Doc Brown from the _Back to the Future_ movies? Do you
have an "impossible" engineering project that only someone like
Doc Brown can solve? My resume is at http://www.guymacon.com/

  #5   Report Post  
Old April 1st 04, 07:14 AM
Mikko Nummelin
 
Posts: n/a
Default Endgame tablebase observation--not always harder

On Thu, 31 Mar 2004, Robert Richter wrote:

Anyway, how often does a computer refuse a devastating move because it
senses that you have one and only one way out, and the way out would
put it behind a pawn?


Best computers are often played against each other and in tournaments
against grandmasters. They cannot afford to make such "devastating moves"
which turn out actually be positional mistakes or even blunders as (at
least) high-level computer opponents and GMs are likely to find out the
"solution", even if there were only one difficult way out. Against lesser
players ... well they anyway make mistakes and give a good computer
program more than enough chances to win.


Mikko Nummelin


  #6   Report Post  
Old April 2nd 04, 09:58 AM
Akorps666
 
Posts: n/a
Default Endgame tablebase observation--not always harder

One modification that might be tried, is for the computers to access both the
Ken Thompson and the Nalimov tablebases, and not just choose the move that
delays the mate the longest, but give some weight to choosing the move that
delays reduction of material the longest.

This only makes sense in computer vs human competition of course

Another trick is to give more weight to moves and sequences of moves that
require "only moves" in response, or have a minimal number of correct reponses,
to force the human to play nearly perfect chess to succeed

Reply
Thread Tools
Display Modes

Posting Rules

Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
endgame tablebase bug/question Thomas rec.games.chess.computer (Computer Chess) 4 November 22nd 03 12:37 PM
Fritz 8.0 and tablebase jar rec.games.chess.computer (Computer Chess) 1 November 14th 03 01:33 PM
Fritz8 EG tablebase drummerman rec.games.chess.computer (Computer Chess) 3 October 28th 03 05:38 AM
Who was better in this endgame? gec rec.games.chess.analysis (Chess Analysis) 3 October 15th 03 06:20 PM
tip: how to improve tablebase access with bigRAM Euc1id rec.games.chess.computer (Computer Chess) 3 October 1st 03 08:13 PM


All times are GMT +1. The time now is 01:48 AM.

Powered by vBulletin® Copyright ©2000 - 2019, Jelsoft Enterprises Ltd.
Copyright 2004-2019 ChessBanter.
The comments are property of their posters.
 

About Us

"It's about Chess"

 

Copyright © 2017