Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Have You seen this: quite unbelievable!

Author: Vincent Diepeveen

Date: 12:11:39 02/04/01

Go up one level in this thread


On February 03, 2001 at 01:52:02, Sandro Necchi wrote:

>On February 02, 2001 at 21:24:43, Vincent Diepeveen wrote:
>
>>On February 02, 2001 at 02:14:41, Sandro Necchi wrote:
>>
>>So in short you prefer a 30 ply material only search
>>with tactical extensions up to say 60 ply or so?
>
>No I meant that if you see a depht of 14/30 it does not mean 14 ply full depht
>and 30 selective, but something like:

>6 ply full depht up to 14 ply selective and up to 30 ply max allowed extension
>search.
>This means that still there is an horizon problem in some cases and only a
>deeper search will let the program see deeply enough:

What program are you referring to here now?

Because if i show depth=14 in diep then some lines are seen up to 128 ply
(usually several tactical lines reach depths like 66 or 72 ply),

so that would be in reality 14/72 then, but for sure it doesn't miss
anything which you mean within that depth in practical game play (of
course zugzwangs in far endgame you can discuss about it needs a
few ply more as normal, but that's a theoretical discussion not
a practical decision).

Perhaps what you mean to say is: you don't see the positional
advantage from preprocessors, you find they play positional moves
based upon a few ply depth instead of the depth they show?

If that's your statement then i have to agree with you.

>30/80 would mean (to me):
>
>12 ply full search up to 30 ply selective and 80 ply max allowed extended
>search. Deeply enough to see what needs to be seen.
>To have such a power needs a much faster computer than today's.
>I do not rely on deep search as I think chess knowledge increase is better, but
>sometimes only a deep search can correct incorrect evaluations/missing chess
>info.

A preprocessor at 30 ply is not going to answer your prays either!

>I hope I was clear enough.
>
>Ciao
>
>Sandro
>>
>>You prefer this over a say 12 ply search of a chessprogram?
>>
>>>On February 01, 2001 at 09:44:28, Vincent Diepeveen wrote:
>>>
>>>>On February 01, 2001 at 04:11:36, Sandro Necchi wrote:
>>>>
>>>>>On February 01, 2001 at 03:48:09, Uri Blass wrote:
>>>>>
>>>>>Dear Uri,
>>>>>
>>>>>>On February 01, 2001 at 02:20:33, Sandro Necchi wrote:
>>>>>>
>>>>>>
>>>>>><snipped>
>>>>>>>No, maybe I did not clearly explain what I meant. I said that a friend of mine,
>>>>>>>testing the program regardless of the opening book, said that it was 100 points
>>>>>>>better than Fritz 6a. So it was not the book making it stronger, but the engine
>>>>>>>being stronger.
>>>>>>
>>>>>>My impression about the engine is different.
>>>>>>
>>>>>>I used ahredder5 for a long time to analyze positions from my correspondence
>>>>>>games when I was in theory.
>>>>>>If the engine is 100 elo better than I expect it to find good moves in the
>>>>>>opening without opwning book.
>>>>>
>>>>>I think that to expect PCs chess programs "to make theory" is wrong. The horizon
>>>>>effect should create a lot of problems on this matter. it really depends on the
>>>>>positions. In some program A should do better than program B and viceversa.
>>>>
>>>>Sorry to fall into this discussion but the word horizon effect
>>>>is no longer valid. If you get a real small search depth like 6 or 8 ply
>>>>then you might refer to horizon effect, but we talk about depths
>>>>of like 13 ply and more here to start with knowing Uri that's the
>>>>minimum depth he let the progs analyze. I'm sure Shredder at
>>>>so many hours a move gets a bit more as 13 also.
>>>>
>>>>Horizon effect is no longer valid then, as that would mean that
>>>>you assume everything is horizon effect till the game is solved by search
>>>>which is quite hard on todays pc's
>>>
>>>No, I do not agree. Well, since the best move is selected by a numerical plus
>>>compared to other moves and the programs are really selective, the depht are not
>>>really valid because a lot of moves are discharged to reach such a dephts. So
>>>some moves are not analized deeply enough and this would done only at high
>>>dephts.
>>>So to really see many moves ahead as it really needs in the early stage of a
>>>game it takes a huge amount of time.
>>>This situation will get better when the full search will reach 30 ply with 60
>>>ply selective.
>>>I think the computer speed must be increased quite a bit to do so, so maybe in
>>>the next 10 years or so.
>>>I agree that a lack of knowledge would effect the search also, but generally
>>>speaking in chess there are so many exceptions that needs to be analised deeper
>>>to see everything clear enough to correct knowledge missing and/or wrong
>>>evaluations.
>>>So, my concept is the programs do not know what they know and somebody must take
>>>a look to them to let them take the correct way.
>>>
>>>Sandro
>>>>
>>>>More valid is the term preprocessor + lack of knowledge.
>>>>
>>>>>>
>>>>>>I found that shredder suggested me a stupid sacrifice and only after many hours
>>>>>>the score went down and it converge to the theory move Re1.
>>>>>>
>>>>>>Here is the opening in my correspondence game against yoav dothan(I am white)
>>>>>>
>>>>>>1.e4 c5 2.Nf3 e6 3.d4 cxd4 4.Nxd4 a6 5.Nc3
>>>>>
>>>>>Even if this is probably the best move, I think for Shredder 5. Bd3 should be
>>>>>better after some moves. I know that the program evaulation is better for Nc3,
>>>>>but it is important the program position understanding after some moves...
>>>>>
>>>>>Qc7 6.Be2 b5 7.0-0 Bb7
>>>>>>
>>>>>>I gave Shredder5 to analyze this position for a long time and it needed many
>>>>>>hours to avoid the bad sacrifice 8.e5
>>>>>>It changed it's mind to Re1 that is the theory move only after many hours.
>>>>>
>>>>>This confirms what I said before. if the position is understood by the program
>>>>>the program would find the correct moves easily.
>>>>>>
>>>>>>I analyzed 8.Re1 when there is an interesting line 8.Re1 b4 9.Nd5.
>>>>>>I can also sacrifice a pawn by 9.Na4
>>>>>>
>>>>>>I did not get a clear consequence about the sacrifices and I decided after a
>>>>>>long think to play 8.a3 and not to sacrifice a pawn or a knight(the game
>>>>>>continued 8...Nf6 9.Qd3 d6 10.Bg5 and it is yoav to move).
>>>>>>
>>>>>>One of the reason that I decided not to sacrifice is the fact that my opponent
>>>>>>does not have to accept and I assume in my correspondence games that my opponent
>>>>>>will probably play the best move.
>>>>>
>>>>>Yes, you are correct.
>>>>>>
>>>>>>It means that if I evaluate that b4 has 50% chance to win and 50% chance to lose
>>>>>>my chances are smaller than 50% because I assume that my opponent will play b4
>>>>>>with probability of more than 50% if it wins and will not play it with
>>>>>>probability of more than 50% if it loses.
>>>>>
>>>>>I agree, but what is best to play is also what will bring you to a position
>>>>>which will suite your style or positions where you can play at your top strenght
>>>>>and not only a general % score.
>>>>>I understand this is not easy to explain and this is why one never stops to
>>>>>learn...
>>>>>>
>>>>>>Uri
>>>>>
>>>>>Sandro



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.