Author: Dan Kiski
Date: 08:15:29 12/15/98
Go up one level in this thread
On December 15, 1998 at 10:29:39, Harald Faber wrote: >On December 15, 1998 at 09:59:41, Dan Kiski wrote: > >>While I agree as I said a programmer can adjust his program for any position, > >Or vice versa! > >>to my knowledge none have done so to soley win the Nunn positions. > >Noone will confirm he did! > >>As both programs are given white and black and as I have said before I play a >>series of games then the program can learn from it's mistakes of the previous > >What if there are positions the program does not "understand"? Then it makes no >difference in paying white or black. > >>game. As I have also said and you have too, IMHO all programs in the top ten of >>the SSDF are relatively equal within say 50 ELO points. Some are naturally >>better than others from certain positions be those positions tactical, >>positional, defensive or attacking, so I don't think any programs are no good. > >Right. It is a point of interpretation if one tries to see differences after >playing the Nunn test. You must admit that in the 10 positions there are some no >program would ever play voluntarily. From what I recall none of the programs will choose the positions as either best line for black or from white. They will all deviate to another line prior to move 10 for black, although this does somewhat depend on which opening book selected and how the opening book is set, example in Fritz 5, if set to play only the strongest opening moves will not play any of these lines, yet if set to play more random lines will play some of them. Basically the engines have to be feed the Nunn positions or in the case of Fritz make a new opening tree using them. >The one and only thing you get out of the Nunn test is which program handles >these 10 opening positions best. Not more, not less. While I generally agree, I still think it is better than letting the engine select at random which line it wants when opening theory can give it a lead. > >>I just do not see any point in playing 50 random games whereby the computer >>engine gets to pick it's best line and comes out of the opening 0.75 up. >>I read once here that Fritz 5.16 has never came out of book with an assesment of >>worse than -0.5, well I have seen cases where computers come out of their >>opening books worse than -1.00 that is why I use the Nunn positions. > >-1.00 or worse indeed often led to a loss. But coming back to your learning >point it will also be interesting to see how the learner reacts after such >opening lines. :-) >And that is why I play a complete series 1x60 and not splitted 5x12 to see the >learning effect.
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.