Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Better usage of EGTBs

Author: Robert Hyatt

Date: 08:05:54 03/26/03

Go up one level in this thread


On March 25, 2003 at 16:41:35, Mike Hood wrote:

>I don't think that Chessbase's tablebase engine (which is activated when a
>tablebase position is on the board) is used optimally in drawn positions. In
>engine-engine matches it makes no difference, but a different strategy could
>give better results against human opponents. Let me explain what I mean, using
>the following position:
>

Crafty has been doing exactly this for many years.  It is the "swindle mode"
that several others now have as an option.



>[D] 8/8/8/7p/2K4P/5k2/8/8 w - - 0 1
>
>Let's assume that the computer is playing White and the human is playing Black.
>The tablebase engine first does a lookup and recognizes that the position is a
>draw; then it finds that there are are four "correct" moves that lead to further
>drawn positions, namely:
>
>1. Kc3
>2. Kd3
>3. Kd4
>4. Kd5
>
>The tablebase engine chooses the first move in the list, Kc3. If the human
>replies with Kf2 the new list of moves is:
>
>1. Kc4
>2. Kc2
>3. Kd2
>4. Kd3
>5. Kd4
>
>The computer will play Kc4, and if the human replies Kf3 we're back at the start
>position again, and the computer will play Kc3 again, heading for a rapid draw
>by threefold repetition.
>
>Not a mistake? Well, be honest, if you were playing this position as White in
>your chess club, and you had 20 minutes on your clock while your opponent has 3
>seconds, would you want a draw by repetition? Of course not.
>
>A better strategy for tablebase engines would be to avoid draws by repetition if
>at all possible. The algorithm is easy to program: if in a drawn EGTB position
>several moves are possible, only choose the first move in the list if it doesn't
>lead to a position that has already been on the board. Otherwise make the next
>move in the list, etc. This will lead to draws under the 50-move rule rather
>than by threefold repetition.
>
>I chose a very simple example to illustrate this point, but look at the example
>in http://www.talkchess.com/forums/1/message.html?290545
>
>[D]4K3/7P/6k1/8/8/8/1q6/7Q w - - 0 102
>
>It would be foolish for White to accept a quick draw by repetition in a KQQKQ
>endgame position.



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.