Author: Uri Blass
Date: 09:06:42 06/04/01
Go up one level in this thread
On June 04, 2001 at 10:56:47, Robert Hyatt wrote:
>On June 04, 2001 at 10:42:17, Uri Blass wrote:
>
>>On June 04, 2001 at 10:17:35, Uri Blass wrote:
>>
>>>On June 04, 2001 at 09:33:39, Sune Larsson wrote:
>>>
>>>>
>>>> [D]2br3r/1pN1pk2/p1np1pq1/6p1/2P1P1p1/1P2Q1B1/P4PPP/3RR1K1 b - - 0 22
>>>>
>>>>
>>>> This is the position from the CCT3 tournament game Insomniac-Shredder.
>>>> After recieving some mails requesting for comments to my game vs Deep Fritz,
>>>> I post this follow up. It focuses on the subjects "doing nothing" vs
>>>> "doing something", how good are top programs in detecting attacks?,
>>>> how intense and strong is their play when examined under a microscope?
>>>> Kramnik is just around the corner...and what could we really expect from
>>>> those coming games?
>>>>
>>>> In this special game, white has just played 22.Rfe1. Shredder didn't like
>>>> the position (some people claimed it was almost lost), and chose "doing
>>>> nothing" with 22.-Rhf8. White then won. But what would happen if we tried
>>>> a more aggressive attacking plan for black? I wanted to know and that's the
>>>> reason for the following game vs Deep Fritz.
>>>>
>>>> I refrain from more conclusions than those embedded in the pgn.game comments.
>>>>
>>>>
>>>>[Event "Nivå=120'/40+60'/20+30'. "]
>>>>[Site "KIL"]
>>>>[Date "2001.06.01"]
>>>>[Round "?"]
>>>>[White "Deep Fritz"]
>>>>[Black "Larsson"]
>>>>[Result "0-1"]
>>>>[SetUp "1"]
>>>>[FEN "2br3r/1pN1pk2/p1np1pq1/6p1/2P1P1p1/1P2Q1B1/P4PPP/3RR1K1 b - - 0 22"]
>>>>[PlyCount "61"]
>>>>[EventDate "2001.06.01"]
>>>>
>>>>{128MB, PowerBig.ctg. PIII 800 In this position Shredder chose a standstill
>>>>and with 22.-Rhf8 intended to do nothing. White eventually broke through with
>>>>c5 and won the game. Black's major problem is his king, but if white is not
>>>>active enough, black can create dangerous counterplay on the kingside. The
>>>>vulnarable squares around white's king are h2 and f3, so a setup with Ne5, Qh5
>>>>(to support a possible Nf3+) and doubling the rooks on the h-file, looks
>>>>promising.} 22... Qh5 {0 Gambit Tiger chooses 22.-e5 with the plan of planting
>>>>the knight on d4. But white gets a dangerous attack after 23.Qb6 Nd4 24.Rxd4
>>>>and later breaking in the center. After some moves GT showed an eval of more
>>>>than -2.00. In this variation black's king is too naked. Instead 22.-Qh5 focus
>>>>on both h2 and f3, and if white is passive, he can easily get in danger.} 23.
>>>>a3 {1.44/14 196 White is doing nothing. This move simply wastes a tempo and
>>>>has no purpose at all. Much better was f.e.23.c5} 23... Rh6 {54 Preparing to
>>>>double the rooks on the h-file, and at the same time setting a little trap for
>>>>white, who now can go pawnhunting.} 24. Qb6 {1.97/14 93 Yes, white sees the
>>>>option of getting some on the Q-side. Much more direct was again 24.c5} 24...
>>>>Ne5 {56 And now black has reached his ideal attacking setup. White has to be
>>>>very careful about the option Nf3+, since there are loads of mate patterns
>>>>with a black pawn on f3.} 25. Nxa6 {2.09/13 75 White takes the pawn and feels
>>>>happy with +2.09. Again white could play 25.c5
>>>
>>>I believe that the time management of Deep Fritz is the main mistake
>>>here.
>>>
>>>Deep Fritz has a tendency to use less time when the score goes up and
>>>it is a clear mistake in the time management.
>>>
>>>This mistake in time management is typical to Fritz's versions.
>>>
>>>If you give Deep Fritz more time to get into depth 14
>>>it can find 25.Ne6 with no problem and I believe that white
>>>has no problem after 25.Ne6.
>>>
>>>Uri
>>
>>I can add that the main problem is
>>not failing high but the fact that Fritz does not like to calculate
>>when it see a big advantage for itself.
>>
>>At depth 14 Deep Fritz fails low but Fritz cannot get into depth 14
>>at tournament time control because it does not like to calculate
>>when the evaluation is more than +2.
>>
>>Uri
>
>
>If that is true, it is a stupid design decision. It is too late to wait
>until the score gets to -2 before you start using a normal amount of time.
>
>I don't use the score to set my target search time. I think it is asking for
>trouble.
The problem is not a big problem because usually when Deep Fritz shows
more than +2 it wins the game but there are cases when it is a problem.
I am not sure about the decision to play faster when the score is more
than +2 but I am sure that I watched Fritz playing faster when
it had clearly a winning position inspite of the fact that
it did not see a forced mate.
In this case Fritz did not guess correctly the right move
so I do not see a logical reason to stop searching after only 75
seconds when you remember that it is 2 hours/40 moves and fritz had
more than average of 3 minutes per move.
Uri
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.