Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Juicy scandal

Author: Jürgen Hartmann

Date: 06:39:17 12/17/98

Go up one level in this thread


On December 17, 1998 at 08:20:58, Robert Hyatt wrote:

>On December 17, 1998 at 07:13:52, Ralph E. Carter wrote:
>
>>Hey everybody. There's a juicy scandal brewing over on rgcc.
>>See Robert Hyatt's replies to the latest post of Amir Ban.
>>
>>Subject: Re: Fritz 5.32: How disruptive to other engines?
>
>
>I don't think it is a "juicy scandal."  I simply asked some questions and
>also pointed out that giving Crafty a "new" command after each move and then
>resetting the board for the next move is really going to defeat a lot of code
>I have developed, and it hurts significantly.
>
>I'd be happy to produce some output showing this if anyone is interested.  Now
>the only question is exactly when/where is this stuff being done?  IE only in
>the winboard engine interface they released, or in the normal fritz 5 engine
>interface as well.
>
>I'm somewhat unhappy with the fact that when they did the modifications to
>crafty to make it work as an engine, they also removed the logging facility so
>that now it is difficult to debug end-user problems and impossible to verify
>what is going on with this latest revelation...
>
>This has also been discussed here (It was first revealed here by Heiko in fact).

There is an interesting engine in Fritz5.32, the Doctor? 3.0. It is not strong
but it learns during analysis. You can show it an incredible difficult
combination, then go back to the beginning position and it will very quickly
find it 'again'. I use this often in my analysis experiments because you can
show the engine dangerous continuations which it wouldn't understand otherwise
and make it very efficiently look for better alternatives. It is a beautiful
feature.

I vaguely remember that I read this is done by not clearing the hash tables
between moves. So I don't know what happens with Crafty but here is at least one
Fritz engine which simply keeps its hash tables. Probably its the decision of
the engine programmer, not of the user interface what to do.

Moreover I think one of the Chessbase programmers once wrote in a German chess
magazine why they do the game analysis always backwards which looks weird at
first sight: He said its done to let engines which don't clear hash tables take
advantage of 'knowing' the full game continuation which should be a powerful
enhancement of the analysis results.

Jürgen Hartmann



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.