Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Is it safe to say then that Computers today play 2800 level chess?

Author: Bob Durrett

Date: 07:43:21 11/27/03

Go up one level in this thread


On November 27, 2003 at 10:02:13, stuart taylor wrote:

>On November 27, 2003 at 09:24:35, Bob Durrett wrote:
>
>>On November 27, 2003 at 08:37:47, stuart taylor wrote:
>>
>>>On November 27, 2003 at 08:25:42, Terry McCracken wrote:
>>>
>>>>On November 27, 2003 at 04:36:20, ALI MIRAFZALI wrote:
>>>>
>>>>>On November 27, 2003 at 03:06:02, Kurt Utzinger wrote:
>>>>>
>>>>>>On November 27, 2003 at 02:59:36, Gerald Wright wrote:
>>>>>>
>>>>>>>The Top players in the computer chess championship are all capable of drawing or
>>>>>>>winning a match vs Kasparov or them in the top 10.
>>>>>>
>>>>>>     As long as 2200 ELO players can get a lot of draws
>>>>>>     with safe and boring playing style the best comp
>>>>>>     programs do not have more than 2400 Elo.
>>>>>>     Kurt
>>>>>What you are saying Kurt does not make sense at all.A 2400 elo player could
>>>>>not draw Kasaprov under any circumstances.Please check the definition of ELO.
>>>>>Also your claim of 2200 Elo players getting draws is Contrary to my own private
>>>>>testing of many many games against 2200 Elo players.
>>>>
>>>>
>>>>Playing computers are _not_ the same as playing Kasparov or _humans_ whatsoever.
>>>>It is often easy to draw machines as Kurt suggests, winning is far more
>>>>difficult, unless of course you obtain a large "book" advantage, with the White
>>>>pieces.
>>>>
>>>>I've have found different ways to neuter computers, and so have many here who
>>>>buy programmes to play against. It's still even quite possible to bring them
>>>>down with carefully played K-Side attacks. By the time the comp sees it, it's
>>>>too late.
>>>
>>>
>>>In MY experience, it's not enough to do something before the computer realizes
>>>it, but that it is VERY CAREFULLY played and worked out, also!
>>> I've often got into positions where I think there could be a brilliant winning
>>>attack esp. kingside, but don't know exactly which one way will do it, if there
>>>IS one way.
>>>I often try, but it's usually not that one way, or at any rate, atleast one of
>>>the moves I make is not according to it.
>>>S.Taylor
>>
>>There is a way to produce a score sheet, of a game with a chess engine, where
>>the human wins.  Simply play a game.  Then go back to the first mistake and make
>>another move.  Repeat this process as many times as is necessary to get the
>>desired result.  This may not work every time, but it should improve the odds in
>>favor of the Human.  Of course, when you publish the scoresheet, it's best to
>>conveniently forget to mention the take-backs.  : )
>>
>>Although this procedure may look bogus and like "cheating," it may be a good way
>>to find the weaknesses in the chess-playing program.  The final scoresheet
>>should be useful to the engine programmer [and maybe to the opening book maker.]
>> I advocate doing this for the sake of improving chess engines.
>>
>>Bob D.
>>
>>Bob D.
>
>That used to beat all engines, but now, even that's not enough. The human would
>have to do something on the way to a brute force search manually, in order to
>try and prove his "brilliancy".
> If you have another computer running at the same time with the same, or a
>stronger program (hardware?), you can do advanced chess to get the right thing
>done. Atleast that would be a bit easier.
>S.Taylor

I like your "advanced chess" idea but it would be necessary for the human
[better if a GM] to guard against being too much influenced by the other engine
[his partner].  That engine will, necessarily, suggest "computer moves."  If the
human is not wary, he/she will fall into the trap of "buying into" the
computer's suggestions.  It's sort of like a Satan and Eve situation, where the
chess engine plays the role of Satan and the human plays the role of Eve.  If
the human is not careful, he may eat a "poisoned apple."  : )

There is another problem I thought of since posting my bulletin.  The hash table
contents will be unknown to the programmer and that may cause some confusion.
Perhaps a very carefully controlled experiment could provide the missing
information.  It might be best to keep pondering off, or maybe erase the hash
tables before each engine move.  [Unclear to me.]

Bob D.



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.