Author: Stuart Cracraft
Date: 14:01:58 12/20/05
Go up one level in this thread
On December 20, 2005 at 16:47:26, Albert Silver wrote: >On December 20, 2005 at 16:30:11, Stuart Cracraft wrote: > >>On December 20, 2005 at 15:17:25, Alex Schmidt wrote: >> >>>Hello, >>> >>>Vasik Rajlich kindly agreed to give us an interview: >>> >>>http://www.uciengines.de/UCI-Engines/Rybka/vriv/vriv.html >>> >>>Thank you Vasik, >>>Alex >> >>Good interview. Nice to attach a face and a personality with the program's >>author. I wish they had gone into his formal studies more. He studied at MIT? >>Anybody have some detail about that? >> >>The comment about his having learned much, in regards to search, from Fruit, >>is a great clue. So he gets that search, adds some to it, then tacks on >>a monster evaluation for middle-game stuff, and voila, Rybka. > >Learning a number of things from Fruit's search, and just taking at and adding >some to it, aren't exactly the same thing. He does say, "Anyway, if I really had >to give a number - my wild guess is that Rybka would be 20 rating points weaker >had Fruit not appeared." > >I'm not an engine programmer, but I have never gotten the impression one could >simply tack on a monster evaluation to a search, any search, without adding the >equivalent of brakes and handbrakes to the engine's speed. Is Fruit so different >that this can be done at so little cost? > >And if it isn't so different, then his search can't be simply Fruit's plus some, >can it? > Hi Albert, I certainly understand your point. I am not so sure. No one comes out of the blue and creates a monster like Rybka so handily without at least some input, some shoulders on which to build. His giving a nod to Fruit source is important and laudable. In the interview, he commented that Fruit 2.1's source code searching provided important insights hence my question about what is the import of the search code in Fruit 2.1. Sure, brakes and the like are all important. No question about the exceptions angle of the equation when applying a monster evaluation with tactical lookahead within the evaluator, as shown by Uri in his problemist positions. Stuart
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.