Author: Uri Blass
Date: 00:20:00 10/06/05
Go up one level in this thread
On October 06, 2005 at 03:17:15, Lance Perkins wrote: >On October 06, 2005 at 02:37:34, Uri Blass wrote: > >>On October 06, 2005 at 02:13:50, Lance Perkins wrote: >> >>>On October 05, 2005 at 21:03:11, Will Singleton wrote: >>> >>>>A comp played this move against my comp on ICC today, took less than 30 secs. >>>>Can any program find it as fast? >>>> >>>>[d]3rn1k1/2qbrp1p/1p1ppnpQ/p5NP/1PP1P3/P1N4R/4BPP1/3R2K1 w - - bm Rdd3 >>> >>>Need 42 secs... >>> >>>Thinker 5.0c on a 2.4ghz AMD >>> 8 99 145 1347078 h5g6 f7g6 d1d2 >>> 8 105 186 1649244 h3h4 d7c6 e2g4 c7c8 >>> 9 108 848 7140678 f2f4 a5b4 a3b4 d7c6 e2g4 >>>10 96 1787 14663556 h5g6 f7g6 h3h4 d7c6 e2d3 >>>10 99 2037 16294503 h3h4 d7c6 d1d2 a5b4 h5g6 >>>10 162 4246 29903901 d1d3 d6d5 c4d5 >> >>congratulations. >> >>42*2.4/2=50.4 seconds >> >>50.4*32=1608.3 >> >>Thinker is more than 32 times faster than Fruit2.2 that still does not find it >>after more than 35 minutes on my A3000(2ghz) after searching more than 1,300,000 >>Knodes >> >>If you can only be more than 32 times faster than Fruit2.2 in every position >>then thinker may be more than 200 elo better than Fruit2.2 >> >>Note that Fruit2.2's behaviour is interesring. >>At depth 15 it seems to be hypnotized by Rd3 >>It cannot be convinced to play it but also cannot leave analyzing this move. >> >>Here is Fruit2.2's output(after 35 minutes of analysis) >> >>New game, >>3rn1k1/2qbrp1p/1p1ppnpQ/p5NP/1PP1P3/P1N4R/4BPP1/3R2K1 w - - 0 1 >> >>Analysis by Fruit 2.2: >> >>1.hxg6 hxg6 >> +- (1.96) Depth: 1/8 00:00:00 >>1.hxg6 fxg6 2.bxa5 >> +- (1.68) Depth: 2/15 00:00:00 >>1.hxg6 fxg6 2.Rd2 axb4 3.axb4 >> +- (1.68) Depth: 3/15 00:00:00 >>1.hxg6 fxg6 2.Rd2 Bc6 3.bxa5 >> +- (1.48) Depth: 4/15 00:00:00 >>1.bxa5 bxa5 2.hxg6 fxg6 3.Nb5 Bxb5 4.cxb5 >> +- (1.57) Depth: 4/16 00:00:00 >>1.Rd2 axb4 2.axb4 gxh5 3.Bxh5 Qxc4 >> +- (1.63) Depth: 4/17 00:00:00 >>1.Rd2 Bc6 2.hxg6 fxg6 3.b5 Bb7 >> +- (1.59) Depth: 5/23 00:00:00 >>1.Rd2 Bc6 2.Bd1 axb4 3.axb4 Ng7 4.hxg6 hxg6 >> +- (1.46) Depth: 6/24 00:00:00 69kN >>1.Rh4 Bc6 2.Bg4 axb4 3.axb4 Ng7 4.hxg6 hxg6 >> +- (1.53) Depth: 6/26 00:00:00 156kN >>1.b5 Ra8 2.Rdd3 gxh5 3.Rd2 e5 >> +- (1.58) Depth: 6/26 00:00:00 219kN >>1.b5 Rc8 2.hxg6 fxg6 3.Rd2 e5 4.Rh4 Qb7 >> +- (1.48) Depth: 7/26 00:00:01 329kN >>1.b5 Rc8 2.Rh4 Qb7 3.Rd2 e5 4.Nd5 Nxd5 5.Qxh7+ Kf8 6.Rxd5 >> +- (1.45) Depth: 8/30 00:00:01 683kN >>1.Rh4 Qa7 2.b5 Rc8 3.Rd2 e5 4.Nd5 Nxd5 5.Qxh7+ Kf8 6.Rxd5 >> +- (1.49) Depth: 8/30 00:00:01 921kN >>1.Rh4 Qa7 2.b5 Rc8 3.Rd2 Qb7 4.hxg6 fxg6 5.Rh3 e5 >> +- (1.49) Depth: 9/32 00:00:03 1859kN >>1.Rh4 Qa7 2.b5 Rc8 3.Rd2 Qb7 4.a4 e5 5.Nd5 Nxd5 6.Qxh7+ Kf8 7.Rxd5 >> +- (1.45) Depth: 10/40 00:00:06 3295kN >>1.Rh4 Qa7 2.b5 Rc8 3.Rd2 Qb7 4.a4 Rc5 5.hxg6 fxg6 6.Rh3 e5 >> +- (1.41) Depth: 11/40 00:00:12 7239kN >>1.Bd3 axb4 2.axb4 Bc6 3.Bc2 Qc8 4.Bb3 Nc7 5.Rf3 Nce8 6.Ra1 Rc7 >> +- (1.57) Depth: 11/40 00:00:30 17310kN >>1.Bd3 axb4 2.axb4 Bc6 3.Bc2 Qc8 4.Bb3 Nc7 5.Rf3 Nce8 6.Ra1 Rc7 7.hxg6 hxg6 >> +- (1.57) Depth: 12/42 00:00:34 20005kN >>1.Bd3 axb4 2.axb4 Bc6 3.Bc2 Bb7 4.e5 dxe5 5.Rxd8 Qxd8 6.Nxh7 Nxh7 7.hxg6 fxg6 >>8.Qxg6+ Rg7 9.Qxe6+ Kf8 10.Rxh7 Rxg2+ 11.Kf1 >> +- (1.81) Depth: 13/46 00:00:52 31002kN >>1.Bd3 axb4 2.axb4 Bc6 3.Bc2 Bb7 4.e5 dxe5 5.Rxd8 Qxd8 6.Nxh7 Nxh7 7.hxg6 fxg6 >>8.Qxg6+ Rg7 9.Qxe6+ Kf8 10.Rxh7 Rxg2+ 11.Kf1 >> +- (1.81) Depth: 14/50 00:01:10 42884kN >>1.Bd3 axb4 2.axb4 Bc6 3.Bc2 Bb7 4.e5 dxe5 5.Rxd8 Qxd8 6.Nxh7 Nxh7 7.hxg6 fxg6 >>8.Qxg6+ Rg7 9.Qxe6+ Kf8 10.Rxh7 Rxg2+ 11.Kf1 >> +- (1.81) Depth: 15/52 00:01:48 67203kN >> >>(, 06.10.2005) >> >> >>Uri > >I think you got the computation incorrectly. Thinker is not 32 times faster than >Fruit. In the Thinker output, you can see that it only does something like >700Knps, while in your output, Fruit does around 600Knps. > >Thinker does: 4246 29903901, which is 29903901 nodes in 42.46 secs. >Fruit does: 00:01:48 67203kN, which is 67203000 nodes in 108 secs. It is clearly more than 32 times faster not in nodes per second but in finding the move Rd3 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.