Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Chessmaster 8000 test: GambitTiger 2 aggressive

Author: Odd Gunnar Malin

Date: 01:18:59 03/08/02

Go up one level in this thread


On March 08, 2002 at 01:12:24, Harald Faber wrote:

>On March 07, 2002 at 15:00:49, John Merlino wrote:
>
>>On March 07, 2002 at 03:36:15, Harald Faber wrote:
>>
>>>- Time control 40/120+60 (although I still doubt that the King uses this time
>>>control properly)
>>
>>The King will handle this time control just fine. It is the CM8000 GUI that does
>>not support it (although the CM9000 GUI will have sudden death).
>>
>>jm
>
>Hmm. It is stil strange. Yesterday evening I started a game in which CM moved
>within 2-3min the first 3-5 moves after the end of the book and reached a bad
>position which CM lost very quickly. But I have also seen that CM takes up to
>11min for a move. I admit I am a bit confused and not 100% convinced that the
>time management with Gunnar malins adapter or with the King iworks correctly.

Couldn't you enable the log feature in the adapter and send the log to me.
In your sample earlier in this thread the engine have the same color that it get
with the new command (black) so there could be another bug than this black/white
commands that I mention in another thread.
If you want to look for bug yourself you could check that the engine receives
the 'level 40 120 0' command at the start of the game and 'time 720000' or close
to this before the first move it should calculate on.

The adapter have to figur out a level command to send to the engine because this
information is only inderect in the uci-protocol so there is room for bugs here.

I could enable black/white command but as I mention earlier I have to do some
tests so this would be done right when you analyze games too. In normal games
there isn't much to test but when the gui-start to go backwards in the game as
it does when you analyze, it send the hole game to the engine, but I convert
this to send undo commands to the engine so that hashtables are kept to benefit
from this analyzing method. To send new or setboard and feed the needed moves
will most certainly break this benefit for many engines.
There is a propose for this in the protocol but as I understand, this is to fix
some old engines behavior e.g. send white-time-black. It could be that this is
the best method anyway. I could also do an extra 'undo' so the engine would act
like it does in normal games.

Beside this there don't need to be a bug because The King used 3 min to move in
a 120 min/40 move game. (120/40=3) If it was confused the calculation would not
be so different. It still should get this 120/40 level but would be told on the
first move after book that it has only 114 min for the rest of the moves. On the
next move it would be back on track again.

Odd Gunnar
(This 'black/white/time' is on top of my todo list but I have to do some payed
job inbetween so that my child stop crying of hunger :) ).



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.