Author: Robert Hyatt
Date: 08:53:45 10/14/03
Go up one level in this thread
On October 14, 2003 at 10:51:59, Gian-Carlo Pascutto wrote: >On October 14, 2003 at 09:41:16, Robert Hyatt wrote: > >>On October 14, 2003 at 07:43:05, Gian-Carlo Pascutto wrote: >> >>>This problem only exists if you are not able to analyze the weaknesses >>>that your engine has well. >>> >>>Conversely, an engine that is stronger at long time controls today will >>>be stronger in rapid next year and stronger in blitz in 2 years. >>> >>>Whereas an engine that is a good blitzer will just get weaker. >>> >>>-- >>>GCP >> >> >>This is simply not true. One example. Try a null-move R=2 program at >>_very_ fast time controls against a program like the king or Hiarcs that >>doesn't seem to use that. At very shallow depths, R=2 causes lots of >>spectacular search failures by hiding simple tactics. At deeper depths >>this effect is reduced. There are plenty of examples of heuristics that >>work well at shallow depths but fail at deeper depths. There are also >>plenty of examples of heuristics that fail at shallow depths but which >>work well at deep depths. >> >>Parallel search efficiency comes to mind. How efficient are you are 3 plies? >>13 plies? >> >>The idea is well-known. > >Uh, where exactly are you disagreeing with me? "an engine that is a good blitzer will just get weaker" or vice versa... An engine that is a good blitzer due to specific eval terms might well be just as strong at long time controls. An engine that is weaker at longer time controls but stronger at shorter time controls is "cart before horse" in the comparison. The rest "stronger in rapid next year" might or might not be true either, unless you assume everybody remains static. > >-- >GCP
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.