Author: Chris Carson
Date: 05:32:27 08/11/99
Go up one level in this thread
On August 11, 1999 at 04:16:21, Tina Long wrote: >On August 10, 1999 at 03:13:44, Shep wrote: > >>On August 09, 1999 at 18:58:55, Dave Gomboc wrote: >> >> >>>Can I ask how you decided that these were very good settings? Is it mostly >>>based on "feel", or test suites, or tournament games, et cetera? How much >>>computation was performed before deciding on these settings? >> >>Depends on what you consider "deciding". The settings were originally composed >>just based on intuition (or "a wild guess" or "Voodoo", if you want :-). >>Didzis and I have been testing them then to make sure they were not really >>"offbeat". >>Scoring 7/10 against other commercial programs so far and 16/21 in test games >>against other CM personalities seems enough evidence they are "not bad" and >>"probably very good" (and I do not just look at the results, but at the way the >>games were played). >>Note that the main settings have probably been thoroughly tested by Mr Bednorz >>before recommending them; I have just tuned them slightly and improved (?) the >>piece values, just like I did back then with my "CM 5555". >>Maybe it was just luck again that my feeling about these settings was right, who >>knows? :-) >> >>--- >>Shep > >Hi guys, > >In my opinion, >Ideally the settings need to vary as the game progresses. One set of settings >for a whole game is several sets too few. > >Maybe extended selective search is better/worse in early middlegame/late >endgame. > >Maybe CM should play defensively until the cracks appear & then switch to >attacking. > >If CM manages to get a two-pawn advantage then it should desire to exchange >pieces, if it loses a pawn then exchanges are less inviting. > >A human player varies their strategies & tactics as the game progresses. > >The Fritz GUI has taken a tentative step in this direction by allowing different >engines for Strategic & Positional phases of the game. I could have two >Fritz5.32's (with different names) tweaked for Strategic & Positional phases of >the game. > >If I were playing a program against a person or a program, and I was allowed (& >if I knew what I was talking about) I would certainly want to tweak the engine >as the game progressed. > >That's why you have your mum on the sideline at Junior Footy, screaming "get >back" or "go go go go go" or "kick him in the..." I digress. > >Maybe the future chess engines will let you tweak for different aspects of the >game. > >Hi guys, >Tina Long Tina, Well said. I think this would be an improvement over todays program weakness. This would provide "adaptability" for the program, a more human trait. I think the settings should be in a file that the program checks, and modifies as the game progresses, thus no human intervention during the game would occur. A program would update the file based on the progression of the game and the programs "plan" for the remainder based of the game. Best Regards, Chris Carson
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.