Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Ferret-Gandalf testposition Re: GambitFruit is busy crushing Spike here

Author: Eelco de Groot

Date: 19:16:27 11/16/05

Go up one level in this thread


On November 16, 2005 at 03:17:53, Bernhard Bauer wrote:

>On November 15, 2005 at 14:43:50, Eelco de Groot wrote:
>
>>On November 15, 2005 at 13:16:27, Marc Lacrosse wrote:
>>
>>>
>>>   Gambitfruit 1.0beta2 vs Spike 1.0a
>>>
>>>   blitz 1 min + 1 sec/move
>>>   no opening book : 50 varied startpositions (Nunn+Noomen)
>>>
>>>   Spike : default parameters
>>>   GF : default + Futility pruning ON + Delta pruning ON
>>>
>>>   partial results : GambitFruit 43.5 - 18.5 Spike
>>>
>>>
>>>
>>>  Pretty impressive !
>>>
>>>  Marc
>>
>>Gambit Fruit seems to gain some "speed" with the futility and delta pruning
>>turned on again:
>>
>>Ferret Gandalf testposition
>>[D]rr4k1/5p2/3p2p1/3P3p/pnP1P2P/Bn2NR2/5PP1/qB1Q2K1 w - - 1 1
>>
>>Gambit Fruit default Beta 2 over Ferret-Gandalf, e5 in 13 minutes, Nf5 2hours
>>
>>00:00:00.0	-0,68	1	3		Bxb4 Rxb4
>>00:00:00.0	-0,69	2	391		Bxb4 Rxb4 Qd3
>>00:00:00.0	0,66	3	2351		Bxb4 Nd4 Qe1 Nxf3+ gxf3
>>00:00:00.0	0,37	4	4394		Bxb4 Nd4 Qe1 Nxf3+ gxf3 a3
>>00:00:00.0	-0,19	5	13334		Bxb4 Nd4 Qe1 Qb2 Rxf7 Kxf7
>>00:00:00.0	-0,32	6	28079		Bxb4 Nd4 Qe1 Qb2 Nc2 Nxf3+ gxf3 a3
>>00:00:00.1	-0,63	7	57569		Bxb4 Nd4 Qe1 Qb2 Nc2 Nxc2 Bxc2 Rxb4
>>00:00:00.2	-0,63	8	118219		Bxb4 Nd4 Qe1 Qb2 Nc2 Nxc2 Bxc2 Rxb4
>>00:00:00.3	-0,87	9	235631		Bxb4 Nd4 Qe1 Qb2 Nc2 Nxc2 Bxc2 Rxb4 Bd3
>>00:00:00.5	-0,97	10	538017		Bxb4 Nd4 Qe1 Qb2 Nc2 Nxc2 Bxc2 Rxb4 Qc3 Qxc3 Rxc3
>>00:00:01.0	-1,22	11	1062815		Bxb4 Nd4 Qe1 Qb2 Nc2 Nxc2 Bxc2 Rxb4 Bd3 a3 Qe3 Qa1+
>>Bf1
>>00:00:02.4	-1,54	12	2377129		Bxb4 Nd4 Qe1 Qb2 Nc2 Nxc2 Bxc2 Rxb4 Rc3 Rc8 Bd3 a3
>>00:00:05.0	-1,91	13	5026777		Bxb4 Nd4 Qe1 Qb2 Nc2 Nxf3+ gxf3 Rxb4 Nxb4 Rb8 Nc6
>>Qxb1 Qf1 Qxf1+ Kxf1 Rb1+ Ke2 Rb2+ Kd3 a3
>>00:00:11.4	-2,13	14	11507575	Bxb4 Nd4 Qe1 Qb2 Nc2 Nxf3+ gxf3 Rxb4 Qxb4 Qxb4 Nxb4
>>Rb8 e5 Rxb4 Bd3 Rb3 Be4 dxe5
>>00:00:33.8	-1,96	15	32068606	Bxb4 Nd4 Qe1 Qb2 Nc2 Nxf3+ gxf3 Rxb4 Qxb4 Qxb4 Nxb4
>>Rb8 Nc2 Rxb1+ Kg2 Rb2 Na3 Kf8 Nb5 Ke7
>>00:01:14.7	-2,66	16	69613966	Bxb4 Nd4 Qe1 Qb2 Bxd6 Qxb1 Qxb1 Rxb1+ Kh2 a3 Be5
>>Nxf3+ gxf3 a2 c5 a1Q Bxa1 Rbxa1 c6 Rc1 Kg3 Ra2
>>00:09:26.0	-2,84	17	501232171	Bxb4 Nd4 Qe1 Qb2 Bxd6 Qxb1 Qxb1 Rxb1+ Kh2 a3 Be5
>>Nxf3+ gxf3 Rc1 d6 a2 d7 f6 Bb2 Kf7 Nd5 a1Q Bxa1 Rcxa1
>>00:13:01.2	-2,79	17	692736310	e5 dxe5 Rg3 Qxa3 Bxg6 Qc1 Bc2+ Kf8 Bxb3 Qxd1+ Bxd1
>>a3 Bb3 a2 Bxa2 Rxa2 Nf5 Rc2 Re3 Rxc4 Rxe5 Nd3
>>00:22:00.4	-2,71	18	1166597250	e5 dxe5 Bxb4 Nd4 Qe1 Nxf3+ gxf3 Qb2 Nc2 a3 Bxa3
>>Qxb1 Qxb1 Rxb1+ Kg2 Rb3 Be7 Ra2 Nb4 Ra4 Nc6 Rxc4 Nxe5
>>00:47:43.4	-2,71	19	2494425208	e5 dxe5 Bxb4 Nd4 Qe1 Nxf3+ gxf3 Qb2 Nc2 a3 Bxa3
>>Qxb1 Qxb1 Rxb1+ Kg2 Rb3 Be7 Ra2 Nb4 Ra4 Nc6 Rxc4 Nxe5
>>02:09:24.9	-2,62	19	2328152205	Nf5 Ra7 Nh6+ Kg7 Nf5+ Kf8 Nxd6 Qxa3 e5 Qa1 Nb5
>>Rxb5 cxb5 Nd4 Rf4 Ne2+ Qxe2 Nxd5 Rc4 Qxb1+ Qf1 Qxf1+ Kxf1 a3 Rc8+ Ke7 Ke2
>>
>>with futility and delta pruning ON e5 in 12 minutes, and considerably less nodes
>>per ply. Fruit 2.2 did this in 25 minutes I believe, but that is about comparing
>>apples and oranges, to stay in fruit terms, it's just an impression
>>I can make a Pro Deo Vulcan version that needs just 6.5 seconds but again that
>>means almost nothing in itself, if it would drop to 25 minutes though I'd worry
>>a bit..
>>
>>Gambit Fruit default Beta 2 with futility and delta pruning
>>
>>00:00:00.0	-0,68	1	3	Bxb4 Rxb4
>>00:00:00.0	-0,69	2	389	Bxb4 Rxb4 Qd3
>>00:00:00.0	0,66	3	2085	Bxb4 Nd4 Qe1 Nxf3+ gxf3
>>00:00:00.0	0,37	4	3163	Bxb4 Nd4 Qe1 Nxf3+ gxf3 a3
>>00:00:00.0	-0,19	5	9344	Bxb4 Nd4 Qe1 Qb2 Rxf7 Kxf7
>>00:00:00.0	-0,32	6	18508	Bxb4 Nd4 Qe1 Qb2 Nc2 Nxf3+ gxf3 a3
>>00:00:00.1	-0,63	7	34326	Bxb4 Nd4 Qe1 Qb2 Nc2 Nxc2 Bxc2 Rxb4
>>00:00:00.1	-0,63	8	69197	Bxb4 Nd4 Qe1 Qb2 Nc2 Nxc2 Bxc2 Rxb4
>>00:00:00.2	-0,87	9	131868	Bxb4 Nd4 Qe1 Qb2 Nc2 Nxc2 Bxc2 Rxb4 Bd3
>>00:00:00.4	-0,97	10	305864	Bxb4 Nd4 Qe1 Qb2 Nc2 Nxc2 Bxc2 Rxb4 Qc3 Qxc3 Rxc3
>>00:00:00.7	-1,22	11	622287	Bxb4 Nd4 Qe1 Qb2 Nc2 Nxc2 Bxc2 Rxb4 Bd3 a3 Qe3 Qa1+
>>Bf1
>>00:00:01.8	-1,54	12	1491254	Bxb4 Nd4 Qe1 Qb2 Nc2 Nxc2 Bxc2 Rxb4 Rc3 Rc8 Bd3 a3
>>00:00:03.6	-1,85	13	3044572	Bxb4 Nd4 Qe1 Qb2 Nc2 Nxf3+ gxf3 Rxb4 Nxb4 Rb8 Nc2
>>Qxb1 Qxb1 Rxb1+ Kg2 Rb2 Nd4
>>00:00:08.0	-2,13	14	6744456	Bxb4 Nd4 Qe1 Qb2 Nc2 Nxf3+ gxf3 Rxb4 Qxb4 Qxb4 Nxb4
>>Rb8 e5 Rxb4 Bd3 Rb3 Be4 dxe5
>>00:00:18.6	-1,96	15	15445922	Bxb4 Nd4 Qe1 Qb2 Nc2 Nxf3+ gxf3 Rxb4 Qxb4 Qxb4 Nxb4
>>Rb8 Nc2 Rxb1+ Kg2 Rb2 Na3 Kf8 Nb5 Ke7
>>00:00:46.1	-2,66	16	38006810	Bxb4 Nd4 Qe1 Qb2 Bxd6 Qxb1 Qxb1 Rxb1+ Kh2 a3 Be5
>>Nxf3+ gxf3 a2 c5 a1Q Bxa1 Rbxa1 c6 Rc1 Kg3 Ra2
>>00:03:55.3	-2,82	17	186178589	Bxb4 Nd4 Qe1 Qb2 Bxd6 Qxb1 Qxb1 Rxb1+ Kh2 a3 Be5
>>Nxf3+ gxf3 Ra5 f4 a2 f5 a1Q Bxa1 Raxa1 fxg6 fxg6 f4 Rb6
>>00:11:58.0	-2,61	17	552805493	e5 dxe5 Bxb4 Nd4 Qe1 Nxf3+ gxf3 Qb2 Nc2 a3 Bxa3
>>Qxb1 Qxb1 Rxb1+ Kg2 Ra4 Bd6 f6 Ne3 Ra2
>>00:20:38.3	-2,71	18	948859856	e5 dxe5 Bxb4 Nd4 Qe1 Nxf3+ gxf3 Qb2 Nc2 a3 Bxa3
>>Qxb1 Qxb1 Rxb1+ Kg2 Rb3 Be7 Ra2 Nb4 Ra4 Nc6 Rxc4 Nxe5
>>00:34:49.0	-2,71	19	1567616135	e5 dxe5 Bxb4 Nd4 Qe1 Nxf3+ gxf3 Qb2 Nc2 a3 Bxa3
>>Qxb1 Qxb1 Rxb1+ Kg2 Rb3 Be7 Ra2 Nb4 Ra4 Nc6 Rxc4 Nxe5
>>
>> Eelco
>
>You didn't mentioned your Hardware.
>For comparison my results.
>On an AMD XP 1800MHz GambitFruit with lots of parameters on found
>e5 at depth 16/54 in 15:57
>e5 at depth 17/54 in 22:11
>Kind regards
>Bernhard

Hello Bernhard,

Finding it at plydepth 16 might be an improvement! 15 minutes compared to 12
minutes on my system is about comparable I think, depending on hashtables, you
probably had more extensions switched on? My results were with 200 MB HT on a
2009 MHz Athlon under Chess Partner interface.

Gambit Fruit default switches to Nf5 after 1 hour 15 I believe, but later
switches to e5 again. Evals for these two moves can vary and probably neither of
them can draw, if Black exchanges Queens and manages to promote the a-pawn, so
in that respect it is not such a good analytical test

Regards,
Eelco



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.