Author: Dieter Buerssner
Date: 15:31:39 09/13/01
Go up one level in this thread
On September 13, 2001 at 17:56:12, Dann Corbit wrote: >Many of the good programs fall for Kb3 (which -- apparently -- is not a good >solution). There are a few good programs which never consider that move. > >I want to understand why some programs choose Kb3. According to my >understanding, the thing that is wrong with Kb3 is that it is basically a time >waster -- it accomplishes no useful purpose. > >What components make a program choose Kb3? >How are some programs able to avoid it? I think, there are a few points. To summarize first: I think, if an engine chooses Kb3 or Be3 at shallow depths is more or less pure luck. If you look at some Be3 PVs you posted, you can see, that they are drawing. Only later, when programs see b6+ in the line, they really see the win. So, for shallow depths, they really do not know which to prefer. I leave one example cited further down. Another point, that makes this position difficult - from the point of view of an evaluation function, that tries to be just a bit clever, but is not really clever enough: White has two protected passed pawns, that are even rather advanced. So, many engines will give those 2 pawns a rather high positional score. But they don't help, one has to sac one (for example by immediate b6 or by b6 later). Here, a slighly clever evaluation function has more problems, to give this, what seems to be a nice advantage, away. Another point is Zugzwang and null move. I think without null move, engines can solve this at smaller depths (but still may need more time). With rather minor changes to Yace, I have seen solved this position in under 10 seconds and in over 2 minutes on my rather slow computer (when I only accept Be3 as a solution, when I see a winning line, which from memory usually means 1. Be3 Bd6 2. Ka5 Bc7+ 3. b6+) Is the version, that solves this fast better? I don't think so. It just happens, that all the pruning and evalution fits better to this position. One example of finding Be3 for wrong reasons first, and for correct reasons later: > 18. 3.29 4 2182591 1. Be3 Bd6 2. Kb3 Be7 3. Kc3 Bd6 4. Kc2 Be7 5. Kd2 >Bd6 6. Kd3 Kb6 7. Kc2 Ka7 8. Bd2 Bc7 9. Kd3 > 19. ++ 20 9005929 1. Be3 > 19. 3.61 66 26557961 1. Be3 Bd6 2. Ka5 Bc7+ 3. b6+ Bxb6+ 4. Kb5 Bc7 5. B >xc5+ Ka8 6. Be3 Kb8 7. c5 Ka8 8. c6 Kb8 9. Kc4 Ka8 10. Kb3 Ba5 11. Kc4 > 19. 3.61 67 27029553 1. Be3 Bd6 2. Ka5 Bc7+ 3. b6+ Bxb6+ 4. Kb5 Bc7 5. B >xc5+ Ka8 6. Be3 Kb8 7. c5 Ka8 8. c6 Kb8 9. Kc4 Ka8 10. Kb3 Ba5 11. Kc4 > 20. ++ 90 35750310 1. Be3 Regards, Dieter
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.