Author: Ingo Althofer
Date: 09:24:13 12/09/97
Go up one level in this thread
On December 08, 1997 at 18:15:37, Don Dailey wrote:
>...See below for parts of the original message. ( My "comfortable" brouser again ... )
Original question was: What are good timing strategies when you have
x minutes for the whole game ?
A few ideas:
(i) a pragmatic one: support the introduction of Fischer clocks.
When you have at least t seconds for each new move, a
timing catastrophy will happen seldomly.
(ii) Your proposal to use 1/N(t) of the remaining time where t is
the move counter, will often be to coarse-grained.
More sophisticated would be the following, and it might be
suitable especially for parallel programs: Take one ( or a few )
of your processors aside and let them play auto-games on fast
playing levels, with the current board position as the starting
point. Record, how often these games have this or that length.
From this length distribution you may define your parameter
1/N more appropriately.
Maybe it makes sense to let this processor play with several
rather different engines so that the incest motif may not become
too dominant.
If you have a normal sequential chess program, you may use some
of the "permanent brain" time for such auto-play matches.
Of course, besides the autoplay results you should also take into
account the time situation of the opponent.
(iii) A rather different point:
Some commercial chess programs have an Annotation feature, for
instance the Fritz-Engines do so. In this feature you have as
one parameter the average time per move which is computed. This
means: When the game you want to analyse has 60 moves ( 30 for
White, 30 for Black ), and when the analysis will finally con-
tain 40 moves, then the time used will be approximately
( 60 + 40 ) * your parameter/move.
So the length of the analysing process will depend heavily on
its final product, and it is often not easy to predict how long
the analysis will be.
Interesting would be another annotation mode: Give the total time
for the analysis, and let the program distribute it properly over the
game. This task of distribution is not a trivial one! ( But it may be
easier than "your" timing problem for "game/x". ) Ingo Althoefer.
>So the idea I propose (and maybe others use something like it) is to
>start N at something like 30-35 and slowly decrease it to something
>between 15 and 20. I believe it's basically correct to have used
>the vast majority of your time at move 40-50. It's probably not wise
>to make N too small though since where the game actually ends is always
>in doubt.
>...
> I'm interested in other ideas for better game/X time controls.
>
>Don
This page took 0.01 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.