Author: Vasik Rajlich
Date: 14:29:04 12/22/05
Go up one level in this thread
On December 22, 2005 at 13:28:48, Barry Culp wrote: >I just got my copy of Rybka 1.01 and the first position I tried to >analyze has a possible problem ? I am using the Fritz 9 GUI on a >P4 3.6 ghz with 512mb hash in infinite analysis mode. > >The older version Rybka 1.0 switches from Rh5 (-4.11) to Ke2 (-.80) in 7:16 >at ply 15. > >The newer version Rybka 1.01 stays on Rh5 ply 13 and doesnt make >any more progress after 30 minutes ??? I was able to reproduce >this "problem" a second time. I even tried booting my PC after each analysis. > >Can anyone with Rybka 1.01 reproduce my results ?? > > >[D] r6k/6p1/pq1pB1Pp/2nP1R2/2Q2p2/P1R2P2/2PK3P/1r6 w - - 0 1 > > >Analysis by Rybka 1.0 Preview 32-bit: > >34.Rh5 Nxe6 35.dxe6 Re8 36.Rb3 Qf2+ 37.Qe2 Qxe2+ 38.Kxe2 Rg1 39.Kd3 Rxg6 > ³ (-0.30) Depth: 11 00:00:13 687kN >34.Rh5 Nxe6 35.dxe6 Re8 36.Rb3 Qf2+ 37.Qe2 Qxe2+ 38.Kxe2 Rg1 39.Kd3 Rxg6 40.Rf5 > ³ (-0.33) Depth: 12 00:00:25 1354kN >34.Rh5 Nxe6 35.dxe6 Re8 36.Rb3 Qf2+ 37.Qe2 Qxe2+ 38.Kxe2 Rg1 39.Kd3 Rxg6 40.Rf5 >Rgxe6 > ³ (-0.33) Depth: 13 00:00:46 2558kN >34.Rh5 Nxe6 35.dxe6 Re8 36.Rf5 Qg1 37.Qe2 Qxg6 38.Rxf4 Rxe6 39.Rc8+ Kh7 40.Qd3 > ³ (-0.48) Depth: 14 00:01:29 5272kN >34.Rh5 Nxe6 35.dxe6 Qf2+ 36.Qe2 Qd4+ 37.Rd3 Qg1 38.Rc3 Rf1 39.Ra5 Rf2 40.Rxa6 >Rxe2+ > -+ (-4.11) Depth: 15 00:04:10 14311kN >34.Ke2 Re8 35.Qxf4 Nxe6 36.dxe6 Rxe6+ 37.Kd2 Qf2+ 38.Kd3 Qe2+ 39.Kd4 Rd1+ 40.Rd3 >Rxd3+ > µ (-0.80) Depth: 15 00:07:16 23074kN > > >Analysis by Rybka 1.01 Preview 2 32-bit: > >34.Rh5 Nxe6 35.dxe6 Re8 36.Rb3 Rxb3 37.cxb3 Qe3+ 38.Kc2 Rxe6 39.Qc8+ Re8 40.Qb7 > ³ (-0.41) Depth: 11 00:00:11 1588kN >34.Rh5 Nxe6 35.dxe6 Re8 36.Rb3 Qf2+ 37.Qe2 Qxe2+ 38.Kxe2 Rg1 39.Rf5 Rxg6 40.Rxf4 > ³ (-0.33) Depth: 12 00:00:24 3348kN >34.Rh5 Nxe6 35.dxe6 Re8 36.Rb3 Qf2+ 37.Qe2 Qxe2+ 38.Kxe2 Rg1 39.Rf5 Rxg6 40.Rxf4 >Rgxe6+ > ³ (-0.40) Depth: 13 00:00:45 6395kN > >No further analysis here after 30 minutes ?? Notice the difference >in node count on plys 11-13 between the 2 versions This has a very simple technical explanation. Rybka 1.0 Beta never resolved mates - all mates were equal. Rybka 1.01 Preview 2 resolves mates under the following two scenarios: 1) Infinite analysis mode 2) When it's time to go mate the opponent (ie. previous search was mating) Currently, Rybka resolves mates very inefficiently. There are a number of solutions, I will improve it in time for Rybka 1.2. For now, Rybka can take a long time to resolve the mate. Please note that this is relevant even though Rh5 doesn't get completely mated. It's obvious there are a lot of mating variations there. As for Aaron's finding - "go depth 999" is not considered infinite analysis mode, so mates are not resolved there. Vas
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.