Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Which new programs are due soon, or are expected to be strongest yet

Author: maria clara benedicto

Date: 18:38:46 03/17/02

Go up one level in this thread


Dear Mr. Christophe Theron,

Sir,

you explanations very fine.

understand clearly.

thanks

maria

p.s.

at least you dont tell us "read ------.doc...."


On March 16, 2002 at 22:34:46, Christophe Theron wrote:

>On March 16, 2002 at 19:53:13, stuart taylor wrote:
>
>>On March 16, 2002 at 10:31:31, Christophe Theron wrote:
>>
>>>On March 15, 2002 at 16:40:55, Uri Blass wrote:
>>>
>>>>On March 15, 2002 at 13:50:33, Christophe Theron wrote:
>>>>
>>>>>On March 15, 2002 at 02:38:05, stuart taylor wrote:
>>>>>
>>>>>>On March 14, 2002 at 08:03:23, Gian-Carlo Pascutto wrote:
>
>>>>>Well I'm glad that you ask. Usually people just ask about the release date of
>>>>>the program, without apparent interest for what will be in it.
>>>>>
>>>>>What's going on with Tiger since version 14.0 is that unlike the previous years
>>>>>I have worked a lot of the EVALUATION of the program. I have tried to work on
>>>>>evaluation holes (mising knowledge about positional features) and tried to fill
>>>>>these holes with good chess knowledge.
>>>>>
>>>>>For example there is much more endgame evaluation in Tiger 15. Also better
>>>>>evaluation of the bishop pair (it took approximately 2 months of extensive
>>>>>research).
>>>>>
>>>>>I have worked on search algorithms less than the previous years because as I
>>>>>said above I focused on the evaluation function. However Tiger15's search is
>>>>>already better than Tiger14's. I'm getting a much better search depth in the
>>>>>endgame for example, and many selection holes have been fixed. I also get better
>>>>>results when mates combinations start to appear on the board.
>>>>>
>>>>>Overall I have the feeling that Chess Tiger 15 will not be much better than CT14
>>>>>in fast games, but that the difference will be clearly seen at long time
>>>>>controls because of better stability (reliability) of the evaluation and less
>>>>>selection holes.
>>>>
>>>>I do not understand
>>>>I remember that you say that fixing the evaluation is more important for palm
>>>>tiger because the difference is bigger when you cannot search deep enough so
>>>>I thought that there is going to be bigger improvement in blitz based on your
>>>>posts.
>>>>
>>>>Uri
>>>
>>>
>>>
>>>There are different categories of knowledge.
>>>
>>>It looks like when you are really lacking depth (Palm version, searching 5-6
>>>plies in blitz but only 3-4 on some moves) you have to add knowledge aimed at
>>>avoiding tactical traps (better evaluation of pins for example) because you
>>>can't even see them by search. I would call this "tactical knowledge".
>>>
>>>Then there is a stage when your search depth starts to be enough to avoid most
>>>tactics (say 10 plies and up). At that point what you need the most is long term
>>>positional knowledge. The kind of knowledge to tell you that you are going into
>>>a won or lost endgame for example, or that this weakness is going to hurt you
>>>until the end of the game.
>>>
>>>In the new Tiger I have been working more on long term positional knowledge.
>>>
>>>In blitz games tactical abilities are more important. In long games, long term
>>>positional knowledge plays a bigger role, and that's why I expect the gains to
>>>be more obvious at long time controls.
>>>
>>>
>>>
>>>    Christophe
>>
>>That's good!
>>Then you can take a vacation for a few years whilst quicker hardware will
>>automatically make your program stronger and stronger.
>>S.Taylor
>
>
>
>That would be great. :)
>
>But I don't really need a vacation. I love what I do.
>
>
>
>    Christophe



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.