Author: Komputer Korner
Date: 04:09:01 07/31/98
Go up one level in this thread
On July 30, 1998 at 10:40:16, Kim Hvarre wrote: >Just an idea (posted to Ed 3 month ago). What do other programmers think? > >Concept: >Establish a possibility to let the ProgramX play (profile) adaptive or >non-adaptive rel. users request, i.e. a mode where the ProgramX "collects" >information about its opponent and then applies to gameplay. >E.g. tends the opponent to play closed/open positions, to offer qualities, to >handle pawns especially well, to go for short or long castling, to play >aggressive (fast kingattacks) or the opposite. To emphasis bi-sh-ops-/knights, >etc. etc. (Others knows that stuff much better). > >A possible implementation could go like this (schematic): > >Make testsuite: (a "closed" playing-environment of e.g. min. 50 games, where the >user can chose to play or not, i.e. if yes he continues (is bound to) where he >left the last time he played in the suite. The suite of cause is compiled and >stored). If no he just plays ProgramX as ProgramX. > >Do profile: After e.g. min. 50 games ProgramX analyses (or do this consecutive >during playing the suite, e.g. allocating 10% of CPU-time, which is realistic >with to days speed) to > >a) Do profile (characteristics could be identified) and weights to the > knowledgeparameters of Pro-gramX are attached. >b) Profile couldn't be determined with certainty, then: > >1) continue suite or >2) start new or >3) abandon. >This could surely combines with choice of the openingbooks/-lines > >If a, this would leave the user with 3 alternatives: > >1) Play ProgramX "as ProgramX" ;-) >2) Play ProgramX as adaptive ProgramX, i.e. test "own kind of gameplay" against >an opponent that with certainty is much stronger in tactics and perhaps in other >aspects, too. >3) Play ProgramX as non-adaptive, i.e. play against "a worst case" opponent, You >know the statement: "I like closed and quiet positions and tends to get to >them…". > >That kind of intelligence I would love and the perspectives of developing the >profiles, according to new testsuites, based on the users development seems very >attractive to me. A chessprogram, that actual "responds". >But as mentioned - just a thought. > >Best regards, > >kim Would your program X (adaptive version) be playing the MAN or playing the BOARD? -- Komputer Korner
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.