Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Fritz 7.0.0.6 has futility pruning on as default.

Author: Bas Hamstra

Date: 13:05:10 01/27/02

Go up one level in this thread


Hi Ed,

On second sight: DOES Rebel see any material gain? Because sees 3.15 at depth=12
and this is not speculative, it sees a material win. If I am not mistaken
Fritzie had kindof +1.60 at first too, before it sees the real blow. And Fritz
is certainly speculative, giving easily half a pawn for an open file on the
enemy castling formation.

Does speculative work? Well isn't ALL eval speculative? There is hardly anyone
that doesn't give a severe penalty for a totally sucked away castling formation
or a far freepawn. I am pragmatic about it: If you see often big evals melting
away, you have gone too far. If you don't, it works. But I would say that the
succesfull Fritz & Tiger are far more speculative than most, so maybe there is a
lot to discover yet. A problem for me is that big eval bonusses seem to crash
search depth. And suddenly all your pieces are at a1 because you are
outsearched.

Best regards,
Bas.

On January 27, 2002 at 13:28:29, Ed Schröder wrote:

>On January 27, 2002 at 09:43:48, Bas Hamstra wrote:
>
>>I saw it earlier Ed :-)
>
>Really? It must be age related then :)
>
>
>>Seems like Rebel is a different animal nowadays. The
>>King used to be more tactical, Rebel better in long games. But now The King
>>needs 2 minutes for this one. Fritz is close at your time, takes slightly more.
>>So in this position Rebel is King :-)
>>
>>But eh, how do you do it? :-)
>
>I checked Century 3 but no sign of 1.Nf6+ not even 1.g6 comes to mind at 11
>plies. So I assume it must be related to the changes I made in my king safety
>spaghetti code.
>
>Read Thorsten, he wants me to climb the mountain without a rope, in other words
>make Rebel a second CSTAL playing poker chess, brilliant instinct sacrifices
>even those with a high error margin. I don't like the high error margin thought,
>I prefer to move on carefully in this area, in other words when you sacrifice it
>should be right in 9 of the 10 cases.
>
>But I am a believer now of speculative evaluation, together with a good search
>depth that filters most of the errors I want to improve Rebel's playing style
>and strength.
>
>Ask Thorsten for the new paradigm, he will tell you into detail.
>
>Ed
>
>
>
>
>>Best regards,
>>Bas.
>>
>>
>>On January 27, 2002 at 09:09:05, Ed Schröder wrote:
>>
>>>On January 27, 2002 at 06:20:59, Bas Hamstra wrote:
>>>
>>>>[D]r2r1bk1/ppqb1ppp/2p1p3/6P1/2PPN3/PB3Q2/1P3P1P/1K1RR3 w - - Nf6!!
>>>>
>>>>Tao solves this in a couple of minutes (I believe 4 minutes to resolve the fail
>>>>high). But with futility pruning turned on it doesn't. I am curious how it
>>>>affects Fritz solution time. Fritz and CM see it pretty quick, in about 2
>>>>minutes.
>>>
>>>55 seconds for Rebel Century 4 (WB), A-1400, 60 Mb hash.
>>>
>>>Ed
>>>
>>>===========================================
>>>
>>>00:00:00.8	0,84	6	151310	Bc2 Be7 Qh5 g6 Qh6 a5
>>>00:00:01.9	0,76	7	288905	Bc2 Be7 Qh5 g6 Qh6 b5 Nf6+ Bxf6
>>>00:00:01.0	0,76	8	728845	Bc2 Be7 Qh5 g6 Qh6 b5 Nf6+ Bxf6
>>>00:00:03.5	0,67	9	1881667	Bc2 Be7 Qh5 g6 Qh6 Qf4 Re3 Bf8
>>>00:00:12.6	0,73	10	7730501	Bc2 Be7 Re3 b5 Qh5 g6 Qh6 Qf4
>>>00:00:25.5	0,75	11	17509807	Bc2 Be7 Re3 c5
>>>00:00:45.6	0,77	11	31601510	g6 hxg6 Ng5 f6
>>>00:00:55.7	0,77	11	38494175	Nf6+
>>>00:01:20.2	1,61	11	55729659	Nf6+ gxf6 gxf6 Bh6 Re5 Kh8 Qh3
>>>
>>>>Best regards,
>>>>Bas.



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.