Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: CM8000 Time Issues

Author: James T. Walker

Date: 19:12:04 12/01/00

Go up one level in this thread


On December 01, 2000 at 17:10:18, John Merlino wrote:

>On December 01, 2000 at 13:06:11, James T. Walker wrote:
>
>>On December 01, 2000 at 00:22:04, John Merlino wrote:
>>
>>>Here's the scoop for all of you who have noticed/complained/been concerned about
>>>how CM8000 is using its time.
>>>
>>>1) There WAS a bug in the implementation of the "Seconds per Move" time control.
>>>It has already been fixed and will be included in the patch.
>>>
>>>2) There IS a bug in the implementation of the "Moves per Minutes" time control.
>>>The development team is currently attempting to fix it and is certain that a fix
>>>will be in the patch.
>>>
>>>Both of these bugs were on the GUI side and are NOT believed to have anything to
>>>do with Johan's engine.
>>>
>>>3) The minimum setting for the Hourglass time control was raised to 30 seconds.
>>>The old minimum of 5 seconds was just too ludicrous.
>>>
>>>4) The documentation refers to the old "Equal Time" time control which existed
>>>in CM7000 and before. This time control was removed for CM8000 and does not
>>>exist in the program.
>>>
>>>Those of you who have mentioned problems with the Fischer or "Minutes per game"
>>>time controls are mistaken. The development team is 110% confident that these
>>>time controls are accurate.
>>>
>>>Can't wait to hear your comments..... ;-)
>>>
>>>jm
>>
>>Hello John,
>>Of course I believe what you say.  I tend to agree that the Fischer or Mim/game
>>are not a bug.  However I must say that the use of time by ChessMaster on those
>>levels is very conservative.  If I was looking for something wrong I would look
>>at the "Pondering" time.  I think ChessMaster is not giving itself enough credit
>>for correctly predicting the opponents moves. In fact it looks like it's not
>>taking any credit for this.  It will be a very nice program when the
>>exterminators are finished.
>>Jim
>
>What exactly do you mean when you say "not giving itself enough credit for
>correctly predicting opponent's moves"? I would like to be precise before
>bringing this up to the development team, and I'm not sure what you are talking
>about.
>
>jm

Hello John,
I will give you two examples.  They are the last two games played by CM8K today
both at Game/1hour.  In the first game CM8K won and used only 15 minutes for the
entire game.  In the second game CM8K lost and used 17 minutes for the entire
game.  Sometimes it normal for a program to use less time when it is winning and
finds it easy to find winning lines.  The reverse is also sometimes true.  That
is when losing it is usual for it to take longer to try to find a way out of
it's troubles.  So rigt away, CM8K does not display either of these qualities.
This is slightly abnormal but not alarming except when given the fact it's using
only about 25% of it's alloted time for the entire game.  When a program
finishes it's search and moves it ponders the next move.  I'm sure you are aware
of this.  When it ponders the correct move it has saved that much time in
searching it's next move.  It's as if CM8K is counting this time as its own and
not allowing for the fact that this time is not coming off its own clock.  I
just think that to use 25% of the alloted time is putting yourself at a
disadvantage when your opponent is using 90-99% of it's time.  Most programs are
very good at using their time.  CM8K is not.  At first I thougt that maybe CM8K
was only serious about it's time allocation if playing rated games.  So I even
played rated games using Chess Tiger.  It had to beat a couple of lower rated
players to get it's rating up to the point where it was allowed to challenge the
ChessMaster.  But nothing changed.  ChessMaster still used a small portion of
its time in the entire game.  If this were my program I would make it use at
least 80-90% of its time.
Jim



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.