Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Writing a meta-language to describe eval function=about Tempi

Author: margolies,marc

Date: 22:44:57 07/02/03

Go up one level in this thread


Why not use some sort of STEP function triggerred by sets of 3 tempi? When the
step is greater than 1, say, an engine that had 'solid' parameters could be
reset to attack.
I don't get the exponential/logarythmic idea in Betsy. This is a good beginning
but perhaps it skews values too much.



On July 02, 2003 at 19:17:58, Landon Rabern wrote:

>On July 02, 2003 at 18:11:30, Matthew White wrote:
>
>>On July 02, 2003 at 15:04:12, Andrei Fortuna wrote:
>>
>>>On July 02, 2003 at 14:42:36, Matthew White wrote:
>>>
>>>>How do you get a program to develop its pieces if you ignore time? Yes,
>>>>piece-square tables are helpful, but how do you keep a program from chasing a
>>>>piece that it thinks it can exchange, but which will result in letting the
>>>>opponent develop his pieces comfortably and start an attack? I realize that we
>>>>use opening books specifically to avoid having to think about these issues, but
>>>>sometimes when opening books end early, I have seen engines un-develop a piece
>>>>that the book just finished developing!
>>>
>>>I have seen this kind of behaviour in my Freyr sometimes. It's because it
>>>doesn't understand the spirit of the position it finds itself when exiting the
>>>opening book and panics.
>>>
>>>Quickest answer would be opening books and piece squares.
>>>Might work giving a small penalty for undeveloped pieces for the first 10 moves
>>>(as a Master friend advised me) but this "first 10 moves" thing is tricky
>>>because after that you have to clear the hashtable as the rules have slightly
>>>changed and you do not want to find hashentries scored by old rules.
>>>
>>>A refinement would be to select opening book lines suited to the playing style
>>>of your program but that takes lots of work and it needs a very strong player
>>>with opening knowledge.
>>>
>>>Without opening book and piecesquares I guess a set of rules would have to be
>>>established, all centered on developing the pieces to good squares.
>>>
>>>How do you define tempi in a chess program ? I am not sure how I could do that
>>>programatically speaking).
>>>
>>>Andrei
>>
>>Defining tempi is a very difficult question... I am not exactly sure how to do
>>it. The only straightforward way that I can think of doing it in an opening is
>>to count the number of pieces that have been moved from their original squares
>>(or the number of moves remaining to connect the rooks), and to do the same for
>>the opponent. The difference is the time advantage/disadvantage. I believe that
>>most programs ignore time as a factor, but I could be wrong. I think it would be
>>a big step towards improving opening play if we could figure out a way to
>>incorporate tempo, though...
>>
>>Matt
>
>Betsy tries to use a similar measure of tempi in the opening.  The advantage is
>exponential ( 3 tempi is worth about 9 times as much as 1 tempi which is worth a
>centipawn).  I really want to play with this more now and I wish I had VC++
>here.
>
>Landon



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.