Author: Tord Romstad
Date: 09:52:23 08/22/03
Go up one level in this thread
On August 22, 2003 at 12:12:42, Uri Blass wrote: >On August 22, 2003 at 11:56:42, Tord Romstad wrote: > >>50 million nodes is a lot. I think most modern programs (even weak ones) >>should solve this a lot faster, and I'm sure Spector would too if you >>had continued developing it. > >I think that you overestimate other programs. >I believe that a lot of programs need more than 50M nodes Perhaps. But so far, Shredder is the only example we have seen. I still think the majority of modern programs should find this in 5M nodes or less, but I could be wrong. My own program is not very good at finding mates (nor any other type of tactics), and I would expect most other programs to be faster. >Maybe you do not use enough pruning so you can solve it fast. I never prune checks; this seems too dangerous. Perhaps Shredder is using an asymmetric search, and prunes some checking moves at even plies when it considers itself to have a winning advantage? What happens if you make the move Qh5+ and let Shredder analyse the position? I don't have Shredder, but my guess is that it will find the mate very quickly if you first execute the right first move. >I believe that solving it too fast is a mistake because it means that in a lot >of cases you extend bad lines. > >The fact that a program can solve it is not enough and we need other positions >to know if it is better in tactics. Of course. >It is easy to be able to solve some positions that you planned to solve but >to fail in positions that you did not plan to solve them. Absolutely. Tuning the program to perform well in some finite set of tactical test positions is definitely not a good idea. Tord
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.