Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: New Engine Rybka

Author: Thomas Mayer

Date: 19:28:15 12/07/05

Go up one level in this thread


Hi,

> The following is from the read me included with the engine :

[..]

> 3) Tablebase support (I should say that it is only my personal opinion that
> tablebases are not a playing strength feature - others may disagree)
> 4) Proper handling and reporting of mates

> Question 1. Is, with the implementation of 3) and 4) will this fix the
> endgame problems discussed here or will it require more work by Vasik and
> therefore will not be fixed in the official release later this month.

well 3) will fix some obvious problems like not being able to win KPK or KBNK
and such trivial stuff. And it might prevent to get in a draw KPK endgame if a
win is still possible. As several experiments did show the influence of TBs on
the overall strength of an engine is very limited and maybe less then 10 Elo.
(And I mean +10 Elo, but also -10 Elo ! Both is thinkable no experiment was
large enough to give us a secure value about this) In the special case with
Rybka I definitely think that it will be on the + side because it seems that it
does not know much about trivial endgames and therefor will profit from them.

4) will solve severall other mating problems where Rybka reports always +320.00
for mates. See: when all moves lead to a mate, but some in x moves and some in
x+y moves and they are ALL evaluated with the same value, the engine chooses the
move randomly. And when it stumbles then finally in a mate it was more or less
by accident or it does not find any other mate with another move

> Question 2. Once the endgame problems are fixed, will this improve the
> strength of the engine.

It will improve the results for sure. We have seen in this forum several draws
of Rybka which should have been won. I am sure that with 3) and 4) Rybka will
win most of those presented games. I don't know how many % they represent in
total, but it is sure that you will see some rating increasement and no lower
rating in ratinglists. By how much ? Well, it depends on the percentage of games
where the current problems are seen. Most play with automatic resign or
adjudication by the GUI -> So Rybka's mating problem isn't seen THAT often. If
they would all play until mate maybe the percentage of this missed wins would be
higher, no idea...

> Perhaps we will know more nearer time of release.

That's for sure... :) At least 3 days after that...

Greets, Thomas



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.