Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: WB engines should become anti-nerd compatible

Author: Dieter Buerssner

Date: 11:56:29 07/15/01

Go up one level in this thread


On July 15, 2001 at 05:33:59, Mike S. wrote:

I find your message title irritating. While I understand the word nerd, I don't
quite understand what you mean by anti-nerd compatible. But when I read "WB
engines should become ...", this is sounds offending to WB engine authors.

I would suggest, that chess-GUIs, that support WB-engines should become
compatible to the Winboard/Xboard protocol. Not only CB forgets about some more
subtle points (like reporting game results, time left on the opponent clock,
etc.)

I wonder if anybody at Chessbase tested Sparring mode with a WB-engine. I wonder
if I am the only one who sees, that Yace gets only about 20% CPU time, when
playing against Fritz with ponder on, on a single CPU computer. Sure, the last
point may be a Yace specific problem, but I have never seen this under WB, where
the resources seem to be given to both opponents in a fair manner.

I can see, that GUI authors have the desire, of a stateless design (probably the
reason of the well known "new" problem), and that they want to have control over
the engine, so that the engine does never start a search by its own (which WB
engines will normally do, with permanent brain on). But then at least, when
already making extensions to the WB-protocol, they should also report the
current state of the permanent brain setting. I guess, that most WB-engines
prefer to see a continious game instead. It is easy enough to not clear hash
tables. For learning features however, it is much easier to see a continious
game with previous scores. Now, any WB-engine author, that wants to do something
of this has to keep track of many more things. This could most probaably easily
be done in the adapter, giving the work to one author, instead of many. The
adapter could keep track, and in normal cases (no takeback, or jumping between
variants or moves in a game), it could just send the new move. This would make
many more engines work much better under their GUI.

When analyzing games (from back to front), they seem to start the engine in
analyze mode or search mode at random. At least I have not found a pattern, and
have not seen anything documented describing this. For Analyze mode, Yace has
persistent hash tables, which should be of significant advantage for such an
Analysis. However, it will not work at its best with analyze game, because of
their implementation.

>But then again, I would'n have the improvements of the new adapter. Although I
>have no idea what they were anyway.

It would be a nice idea, to document these changes.

>Theoretically, Der Bringer with it's
>own great GUI should benefit from that; unfortunately he's not among the top in
>strength - yet.

From recent results I have seen by Der Bringer, this might have changed.

Regards,
Dieter




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.