Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Correspondence Match (To Uri and Dr. Hyatt)

Author: Steve Coladonato

Date: 06:03:28 05/05/00

Go up one level in this thread


On May 04, 2000 at 22:26:06, Robert Hyatt wrote:

>On May 04, 2000 at 10:31:39, Steve Coladonato wrote:
>
>>On May 04, 2000 at 10:06:28, Robert Hyatt wrote:
>>
>>>On May 04, 2000 at 08:13:51, Steve Coladonato wrote:
>>>
>>>>On May 03, 2000 at 18:26:09, Robert Hyatt wrote:
>>>>
>>>>>On May 03, 2000 at 12:52:01, Steve Coladonato wrote:
>>>>>
>>>>>>On May 03, 2000 at 10:38:57, blass uri wrote:
>>>>>>
>>>>>>>On May 03, 2000 at 09:41:19, Steve Coladonato wrote:
>>>>>>>
>>>>>>>>On May 03, 2000 at 03:29:48, Jouni Uski wrote:
>>>>>>>>
>>>>>>>>>On May 02, 2000 at 13:03:47, Steve Coladonato wrote:
>>>>>>>>>
>>>>>>>>>>There is a correspondence match going on between Steve Ham and both Fritz 6(a)
>>>>>>>>>>and Nimzo 7.32.  The games are documented at the following site:
>>>>>>>>>>
>>>>>>>>>>http://correspondencechess.com/campbell/index.htm
>>>>>>>>>>
>>>>>>>>>>They appear to be quite interesting and the analysis by Mr. Ham is very
>>>>>>>>>>extensive.  It's interesting that even after 19-21 hrs of evaluation, the
>>>>>>>>>>computers are only getting to 15-16 ply.  Also, it looks like Mr. Ham has the
>>>>>>>>>>upper hand in the games.
>>>>>>>>>>
>>>>>>>>>>Regards.
>>>>>>>>>>
>>>>>>>>>>Steve
>>>>>>>>>
>>>>>>>>>I am not so sure if Ham has upper hand. And note, that most moves were played in
>>>>>>>>>3-best move mode!
>>>>>>>>>
>>>>>>>>>Jouni
>>>>>>>>
>>>>>>>>Jouni,
>>>>>>>>
>>>>>>>>What is "3-best move mode"?
>>>>>>>
>>>>>>>Chessbase engines can search the 3 best move instead of only searching for the
>>>>>>>best move.
>>>>>>>
>>>>>>>They did it in the beginning of the game and probably they could search more
>>>>>>>deep by searching only for the best move.
>>>>>>>
>>>>>>>searching for 3 best moves instead of only the best move is about the same as
>>>>>>>being 2-3 times slower.
>>>>>>>
>>>>>>>Even if we do not assume diminishing return from being 2-3 times faster
>>>>>>>the demage for programs in this case is not more than 100 elo and if we consider
>>>>>>>also the fact that the programs did it only in the opening the demage is
>>>>>>>probably less than 50 elo so it will probably not change the reuslt of the match
>>>>>>>because the expected changed in the result is less than 0.25 point
>>>>>>>
>>>>>>>Uri
>>>>>>
>>>>>>Uri,
>>>>>>
>>>>>>When a computer engine evaluates a position, does it not take all the possible
>>>>>>moves and compute an evaluation for each move?  In this case the three "highest"
>>>>>>scores would be the top three and there is really no effect on the processing.
>>>>>>I understand that variations within a given move are also calculated but is this
>>>>>>not just normal processing?
>>>>>>
>>>>>>Steve
>>>>>
>>>>>
>>>>>no.  Alpha/beta finds the best move and only proves that the other moves are
>>>>>worse, without proving how much worse they are.  To do this requires a lot more
>>>>>time.
>>>>
>>>>You have both given me essentially the same answer.  I've never looked at the
>>>>code for a chess engine so I don't know exactly what Alpha/Beta does.  But the
>>>>answers here are confusing to me.  I was under the impression that the best move
>>>>was determined by calculating the eval for the candidate moves.  Your answers
>>>>are implying that that is incorrect and something else is used to determine the
>>>>best move, not the eval for the position.  But if that is the case, is not what
>>>>the program calculates somehow related to the eval?  And if so, saving the
>>>>result in an array would not incur that much more overhead so that the program
>>>>would know what the top three moves are or rather the order of all candidate
>>>>moves based on whatever it is calculating.
>>>>
>>>>Steve
>>>
>>>
>>>Here is the idea:
>>>
>>>You have three holes in the wall.  Your task is this:  "stick your hand in
>>>each hole, for exactly one minute, and then report which one gave you the
>>>most pleasant (or least unpleasant) experience."
>>>
>>>You stick your hand in hole #1.  For one minute, you get nothing but warm
>>>water.  You stick your hand in hole #1.  You are instantly greeted by very
>>>hot water.  This is already more unpleasant than holee #1. Do you wait around
>>>to see if it gets to the boiling point or do you leave _right now_.  I leave
>>>now, as I have already proved that it is worse than #1.  I don't know how much
>>>worse yet (to discover this I have to wait for the full minute).  I go on to
>>>hole #3 and immediately get doused by salt-water at about 30 degrees F.  That
>>>is very cold and much worse than hole number one.  Do I stick around to see if
>>>it gets worse, or do I quit not?
>>>
>>>Total time spent = 1 minute in hole 1, 1 sec in hole 2, one sec in hole 3.  I
>>>only proved that 2 and 3 were worse, but not how much worse.
>>>
>>>That is how alpha/beta works.
>>
>>That makes sense.  But let's start with hole #2 first.  So now I stick my hand
>>in, get doused with very hot water but hang around for a minute.  Now I go to
>>hole #1 and get greeted with warm water.  Nice.  So now the best move is hole #1
>>not hole #2.  But do I hang around in hole #1 for a while or immediately move on
>>to check hole #3.  And what about that shark that's about ready to snatch off my
>>hand should I dally a bit longer in hole #1.
>>
>>It seems like the best move is determined by a very superficial ply movement
>>once something is deemed to be best.  But maybe the best move is actually that
>>knight sac (very cold water) that is rewarded 6 ply later.
>>
>>Of course I can't argue with success.  Am I correct in now believing that ply
>>depth has nothing to do with best move?
>>
>>Steve
>
>
>You fell for my trap.  :)
>
>_now_ you see why move ordering is so critical for good alpha/beta performance.
>If you start with the worst hole, and then the next-worst one, and finally the
>best one, you spend 3 minutes...
>
>If you start with the best, you spend about 62 seconds.
>
>Moral of the story?  Pick the best hole (or move) first, as often as you
>can.  Hence the idea of winning captures, hash move, killer moves, etc.
>
>Bob

Thanks Bob.  Because of this post I found a site that seems to have some good
information about the algorithms that are used in chess software.

http://www.xs4all.nl/~verhelst/chess/programming.html

I'll be checking this site out before I delve into this area chess engines
again.

Steve



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.