Computer Chess Club Archives


Search

Terms

Messages

Subject: S&D 6_7e3 and Ferret-Gandalf

Author: Eelco de Groot

Date: 21:02:13 10/30/04

Go up one level in this thread


On October 30, 2004 at 21:34:28, Albert Silver wrote:

>On October 30, 2004 at 20:47:24, Eelco de Groot wrote:
>
>>Though there is probably no connection with the ply displayproblem from Thomas,
>>I just got a case not of missing lines, but one of duplicate lines with a
>>version of "Search & Destroy" settings, that I call S&D 6_7e. I suspect the
>>parameter [Pruning = MISC_01]* TREE: if (reduce & rd=0) goto tr76
>>to be the culprit. The extra lines at ply ten can be partly explained as false
>>fail highs, but why are there two of them it seems? The position was the one Bas
>>Hamstra gave below from the Tao-Tiger game.
>
>It is caused by the parameter:
>
>[Pruning = MISC_65]             * research fail-high-error one ply deeper (!)
>
>If you turn it off, it won't repeat the line. This way it is actually taking
>those two lines to 11 plies to see whether the extra ply doesn't help resolve
>its indecision.
>
>                                            Albert
>
>>

Thanks Albert, I suppose that makes sense! Seems like a good search-idea from
Ed.

I'm having fun with your Search & Destroy settings. I noticed that you put
almost all the tactical extensions there are in this one, so I presume S&D 6_7
may also be good at tactical problems?

I did also try the parameters for MISC_35 that Ed gave in his post on Thorsten's
forum,

[M37_VAL = 1.00] * 1.00
[M37_count = 2] * use 2 or 3 (default is 4)
[M37_max = 2] * maximum=2

but maybe that is a bit too much, seems to give even more extensions. As the
settings are now, in some positions there can be too many extensions I think,
but I don't know if that hurts game play. It helps if you have enough space for
hashtables. S7D 6_7e3 has some higher mobility -I consider that more a
positional parameter than a tactics change-, then there is the infamous MISC_1
and I was still experimenting with lower Adjusted Selectivity, which as I
understand it gives more "brute force". All these three changes may not be very
good for games though, and lower Adjusted Selectivity certainly has a cost in
nodes. Just experimenting...

-----------------------------------------------------------------------------------
[Personality = Pro Deo S&D67e3.ENG]  * Search & Destroy v6.7 (October 2004) but
slightly modified
[AUTHOR = Albert Silver]
[Pawn Value = 100]
[Knight Value = 100]
[Bishop Value = 100]
[Rook Value = 100]
[Queen Value = 100]
[My Queen = 101]                * new since September 1, 2004
[King Safety = 165]
[Mobility = 125]		* [modified &e3 from 116]
[Pawn Structure = 100]
[Passed Pawns = 111]
[Pins = 105]
[Bishop Pair = 100]
[Chess Knowledge = 100]
[Attractiveness = 103]
[Attacking = 102]
[Strength of Play = 100]
[Draw Contempt Factor = 0.00]
[Selective Search = 100]
[Search Technique = NULLMOVE]

[Engine Learner = off]          * off|on
[Book Learner = off]     	* off|passive|moderate|strong|aggressive
[Position Learner = off]        * off|on
[Extended Book Learner = off]   * off|read|write|read&write
[Clear Hash Table = Always]     * Always|Never|PB only
[Pruning = MISC_01]             * TREE: if (reduce & rd=0) goto tr76 [&e3]
[Pruning = MISC_17]             *
[Pruning = MISC_14]             * Queen ending (new style)
[Pruning = MISC_25]             *
[Pruning = MISC_37]             *
[Pruning = MISC_40]             *
[Pruning = MISC_52]             *
[Pruning = MISC_66]             *
[Pruning = MISC_69]             *
[Pruning = MISC_78]             * adaptive nullmove R=2/3
[Pruning = MISC_03]             * move ordering speedup
[Pruning = MISC_39]
[Pruning = MISC_21]             * Lazy Eval (tuning)
[Checks Depth = 1]              * limit QS checks to 1 ply
[Pruning = MISC_30] 		* extend singular checks
[Pruning = MISC_42] 		* extend first check too
[Pruning = MISC_47] 		* King Safety extension
[Pruning = MISC_26] 		* Endgame extensions
[Pruning = MISC_07]             * No recapture extension when capture is just a
pawn (speedup=13%)
[Pruning = MISC_65]             * research fail-high-error one ply deeper (!)
[Pruning = MISC_28]             * More check extensions (!)
[Pruning = MISC_22]		* Slightly safer Selectivity
[Pruning = MISC_38]             * New capture extension
[Pruning = MISC_44]             * New endgame extension
[Lazyeval MIDG = moderate]       * normal|moderate|strong        promising
*
[Center Control = 25]           * 25
[Bishop Mobility = 25]          * 25
[Right to Move = 75]            * 75
[Strong Squares = 125]          * 125
[Pawn Pressure midg = 165]      * weight pawn pressure middlegame
[Pawn Pressure end0 = 140]      * weight pawn pressure early endgame
[Pawn Pressure end1 = 125]      * weight pawn pressure endgame
[Pawn Pressure end2 = 110]      * weight pawn pressure simple endgame

[Knight Mobility = 88]          * 88
[Double Pawns = 112]            * 150
[Pawn Formation = 108]          * 100
[Weak Pawns = 112]              * 100

[Passed Pawn King Tropism = 100]    * king tropism to own passed pawns
[Progressive Isolated Pawns = 100]  * weight isolated pawns (new style)

[Adjusted Selectivity MIDG = 4] * Maximum selective depth(midgame)[&e3]
[Adjusted Selectivity END0 = 4] * Maximum selective depth (early endgame)[&e3]
[Adjusted Selectivity END1 = 4] * Maximum selective depth (endgame)[&e3]
[Adjusted Selectivity END2 = 3] * Maximum selective depth (simple endgame)
[Time Control = 125]                * 125
[Time Control Status = ALWAYS]      * ALWAYS|PB ONLY
*
[Minimum Knight Mobility = 100] * 100
[Minimum Bishop Mobility = 100] * 100

[ANTI-GM = OFF]
[EVALUATION = SMART]
[Extensions (remaining)= 3]
[Extensions (checks)= 0]
[Extensions (captures)= 0]

--------------------------------------------------------------------------------
---

I like the last result for the Ferret-Gandalf position though, an alternative
for Bxb4 in only 6 minutes 40 seconds! That is worth a few hours precious
nightrest, but you put a lot more time in it than that, let alone Ed. Thanks
guys!

S&D 6_7e3 and Ferret-Gandalf, 6 minutes 40 seconds to play alternative for Bxb4!
(Celeron 500 MHz 28 Mb HT)

[D]rr4k1/5p2/3p2p1/3P3p/pnP1P2P/Bn2NR2/5PP1/qB1Q2K1 w - - 1 1

00:00:04.6	-1,40	1	7	Bxb4 Rxb4
00:00:04.7	-0,16	1	146	Nc2
00:00:04.7	-1,86	2	587	Nc2 Nxc2 Bxd6
00:00:04.7	-1,86	2	824	Bxb4
00:00:04.7	1,53	2	992	Bxb4 Nd4
00:00:04.7	-0,60	3	2050	Bxb4 Nd4 Nc2 Nxc2
00:00:04.8	-0,60	4	2971	Bxb4 Nd4 Nc2 Nxc2
00:00:04.8	-0,41	5	8076	Bxb4 Nd4 Qe1 Nxf3+ gxf3 Qf6
00:00:04.8	-1,32	6	25362	Bxb4 Nd4 Qe1 Qb2 Nc2 Nxc2 Bxc2
00:00:05.1	-1,32	7	75229	Bxb4 Nd4 Qe1 Qb2 Nc2 Nxc2 Bxc2 Rxb4
00:00:06.0	-1,32	8	183501	Bxb4 Nd4 Qe1 Qb2 Nc2 Nxc2 Bxc2 Rxb4
00:00:07.4	-1,14	9	361766	Bxb4 Nd4 Qe1 Qb2 Nc2 Nxc2 Bxc2
00:00:11.1	-1,38	10	1039219	Bxb4 Nd4 Qe1 Qb2 Nc2 Nxc2 Bxc2 Rxb4 c5 dxc5
00:00:19.2	-1,57	11	2713762	Bxb4 Nd4 Qe1 Qb2 Nc2 Nxc2 Bxc2 Rxb4 Bd3 Rab8 Qe3
00:00:38.6	-1,82	12	6284014	Bxb4 Nd4 Qe1 Qb2 Nc2 Nxf3+ gxf3 Rxb4 Qxb4 Qxb4
00:01:40.4	-2,19	13	18360159	Bxb4 Nd4 Qe1 Qb2 Nc2 Nxc2 Bxc2 Rxb4 e5
00:06:40.5	-2,15	13	69282958	Nf5 Qxa3 Nh6+ Kg7 Rxf7+ Kxh6 Qf3
00:26:57.1	-1,89	14	283712648	Nf5 Ra7 Nxd6 Qxa3 Nb5
01:48:37.5	-2,15	15	1139317753	Nf5 Ra7 Nh6+ Kg7 Nf5+ gxf5



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.