Author: Christophe Theron
Date: 23:13:27 10/11/03
Go up one level in this thread
On October 11, 2003 at 16:44:58, Uri Blass wrote:
>On October 11, 2003 at 16:04:04, Christophe Theron wrote:
>
>>On October 11, 2003 at 14:12:41, Uri Blass wrote:
>>
>>>On October 11, 2003 at 14:05:35, Steve Maughan wrote:
>>>
>>>>Christophe,
>>>>
>>>>>1) I have written an extra engine core ("FUGU") which is extremely "tight". It
>>>>>updates an internal representation of the board that is simpler than the one
>>>>>of the classical Tiger engine, and it does not apply all the rules of chess
>>>>>(some rules like en-passant, castling, underpromoting and 50-moves rule are
>>>>>complex to handle and slow the engine down significantly). Doing without some
>>>>>rules in the deepest part of the tree goes mostly unnoticed and help to speed
>>>>>the engine up tremendously.
>>>>
>>>>Now that is an interesting idea. I have heard of top programs that ignore some
>>>>rules for the whole of the search tree (e.g. Junior 5 ignored underpromotion)
>>>>but I haven't heard of anyone ignoring some rules when near the tips of the tree
>>>>(which is most of the search).
>>>>
>>>>Regards,
>>>>
>>>>Steve
>>>
>>>I guess that ignoring the rules is dependent on alpha and beta(otherwise it is
>>>dangerous because the program may see mate based on the assumption that the
>>>opponent cannot play an enpassant capture and only later consider the capture
>>>and find that the mate was an illusion).
>>>
>>>I hope that at least when new tiger announce mate it does not prune enpassent
>>>captures or underpromotions of the opponent.
>>>
>>>Uri
>>
>>
>>
>>This technique can be indeed extremely dangerous. The FUGU engine, by a strict
>>definition, does not play chess!!!
>>
>>But it's not new. As you say, the previous versions of Junior were not playing
>>chess but a slight variant of chess. However it was playing chess brilliantly.
>>
>>I must pay attention to some cases like the one you mention.
>>
>>In this case FUGU would not say it's a checkmate. It would just return a low
>>score (alpha). So far I have never seen any problem because of it.
>>
>>
>>
>> Christophe
>
>Did you test FUGU against tiger15?
>Is it already better than tiger15 based on your tests or do you need more tuning
>for that purpose.
>
>The impression based on the results of it in the french championship suggests
>that at least in that competition it was weaker than tiger15.
>
>Uri
You know very well that you cannot deduce much from a few games.
CT X "FUGU" has some weaknesses that I need to address. In particular using it
in the endgame does not seem like a good idea because of the deficiencies of the
FUGU evaluation of passed pawns.
This stuff simply needs more work...
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.