Author: Vasik Rajlich
Date: 15:34:08 12/25/05
Go up one level in this thread
On December 25, 2005 at 14:22:06, Albert Silver wrote: >On December 25, 2005 at 14:00:34, Vasik Rajlich wrote: > >>On December 25, 2005 at 09:51:18, Sune Larsson wrote: >> >>>> >>>>I've already managed to make a rough design of the endgame knowledge solution >>>>for Rybka 1.2, and this position won't be covered by it. >>>> >>>>In the long term, however, an engine should solve this position as well. It's >>>>part of the bigger concept of knowing when you can't make progress. >>>> >>>>Vas >>> >>> >>> Yes, it's important to know when no progress is possible. Like in this position >>> which comes from a Nunn ending1 match between Fritz9 and Rybka. Rybka entered >>> this position happily, as black, showing + 4.02. But of course it's a draw. >>> The white rook is rock solid on f6. >>> >>> [D]8/4k3/5Rp1/6P1/3K1P2/1P6/8/7q b - - 0 25 >>> >>> >>> /S >> >>Yup. Rybka 1.2 will actually do ok here, but there are some deeper questions. >>One idea I have thought about a little bit is that an advantage should grow over >>time (ie with more search). I think it's true of every serious advantage, but >>the opposite may be true of small advantages. >> >>Vas > > >Here's an oddity I ran accross: > >[D]2k4r/6R1/p1p1P3/8/2P5/P7/1P3r1p/1K5R b - - 0 54 > >For some reason, Rybka can't seem to find 54...Re2 in a reasonable time, >contrary to other engines. It insists on Kd8 which loses quickly. > > Albert At first glance it looks like a search problem. White wins here with 1. .. Kd8 2. Rb7 Rh6 3. e7 Ke8 4. Re1 Re2 5. Rxe2 h1/Q which Rybka isn't going to like for a while. This will all be re-visited later. 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.