Author: Christophe Theron
Date: 10:33:59 05/22/98
Go up one level in this thread
On May 21, 1998 at 13:48:16, Enrique Irazoqui wrote:
>On May 21, 1998 at 13:39:39, Christophe Theron wrote:
>
>>On May 21, 1998 at 13:03:00, Enrique Irazoqui wrote:
>>
>>>On May 21, 1998 at 07:01:18, Georg Langrath wrote:
>>>
>>>>6k1/8/p3r2p/1p1pPpp1/1n1pP3/1Pq4P/PR4PB/3QN1K1 b
>>>>
>>>>This epd-file represents a position between Spassky-Tal 1959. Spassky
>>>>took the rook 1…Dxb2, and who hadn’t. But this is wrong. Even if you
>>>>play white with a weaker chessprogram against black with a stronger one,
>>>>white wins after that move.. The best move is 1…dxe4. On the opposite,
>>>>here a weaker black program always wins over a stronger black program
>>>>after that move.
>>>>
>>>>I wrote about the position in PLY (SSDF) about eight years ago. Even the
>>>>strongest chessprograms didn’t find the right move in those days. Not
>>>>even in analyze for a week or so. Does you program find it? Rebel9 find
>>>>it in analyze, but not in 3 minutes on my computer (Pentium 133).
>>>>
>>>>But Fritz 5 find the right move in about 10 seconds, and doesn’t change
>>>>its mind then! Nearly unbelievable.
>>>>
>>>>I have tested Fritz5 on the selfgoing update, and now I have ordered it.
>>>>Perhaps Chessbase shouldn’t be in a such hurry to put it away, when they
>>>>found that it was possible to make it selfgoing. (You have to put two
>>>>files to it from Cblight only).
>>>>
>>>>Not only I had buyed it after testing it, I think.
>>>>
>>>>Georg
>>>
>>>On a P200MMX:
>>>
>>>Fritz 5 finds dxe4 in 3 seconds (!!!).
>>>Rebel 9, 185 seconds.
>>>Junior 4.6, Hiarcs 6, Mchess 7.1, Nimzo98, Shredder 2, Genius 5 and CST
>>>take over 5 minutes.
>>>
>>>Very nice position. Thanks.
>>>
>>>Enrique
>>
>>Chess Tiger 11.4, K5-100MHz, 16Mb hash, has a fail low on Qxb2 in 0.49s,
>>and finds dxe4 in 0.71s.
>>
>>I let it run for several minutes and it never changed its mind.
>>
>>Just to have some fun, I then ran the position with Tiger running on my
>>old 386sx 20MHz notebook (2Mb hash). It rejects Qxb2 in 18.78s (fail
>>low), and would play dxe4 in 28.34s.
>>
>>So this position is maybe too easy for current programs.
>
>Tiger's time is impressive, but take a look at most top programs...
>
>One of my very favorite positions is the Badai problem I posted last
>Summer:
>8/1p6/p4p1p/2p1P2k/5P2/6pP/1P4P1/6K1 w - - 0 0
>
>White to play 1.f5
>
>How is Tiger doing with this one?
>
>Enrique
Tiger running on K5-100MHz, 32Mb hash, finds 1.f5 in 293.85s (4:54) at
ply 17 (it would find it in 147s = 2:27 on K6-200).
From ply 1 to ply 16, Tiger would play 1.e6. But the score of this move
slowly decreases at each iteration.
At ply 17, in 57.56s, e6 is evaluated at -0.08.
Then Tiger chooses f5, with a fail high. In 293.85s, f5 is evaluated at
+0.88 with the PV:
1. f5 fxe5 2. h4 b5 3. Kf1 e4 4. Ke2 c4 5. Ke3 b4 ...
At ply 18, in 893.75s (14:54), a fail high is detected for f5.
Then in 1260S (21:00), f5 is evaluated +5.62 with the same PV.
I let it run till ply 19, and the same move with the same score and PV
would be played.
How did the other programs on this position?
Christophe
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.