Author: Robert Hyatt
Date: 20:34:04 09/12/98
Go up one level in this thread
On September 12, 1998 at 14:04:40, Serge Desmarais wrote: >On September 12, 1998 at 08:10:14, Moritz Berger wrote: > >>On September 11, 1998 at 17:37:48, Serge Desmarais wrote: >> >>>On September 11, 1998 at 06:25:27, Moritz Berger wrote: >>> >>>> >>>>>You surprise me, here, when you say that Fritz plays opening moves with ONE >>>>>occurence! On my computer, all the moves that are unique get a 0% chance of >>>>>being played. Of course, if all the moves in a given position were played less >>>>>than 10 times, that is another story. Did you set your parameters for Fritz to >>>>>play a greater diversity of moves than the default settings? >>>>> >>>>> >>>>>Serge Desmarais >>>> >>>>Try it yourself: Open tree F11, follow the most popular move until one game >>>>remains, click "move now" button, Fritz will display "in book" in the "seach >>>>info display" and play the book line until the end. >>>> >>>>Moritz >>> >>> But that is another story! If ONLY ONE move is in book for a given position, >>>it has no choice but to play it. But up to there, it must be a move that won the >>>game. Do you think it should start "thinking" when only ONE game is listed for >>>the only move given or for all the available move? That would be an interesting >>>idea. Personnally, I think that Fritz should ADD the first move played by its >>>opponent that is not in its book and also should start thinking if the results >>>obtained by itself when playing all the available moves (or the unique move) >>>weren't good. Of course, not based on blitz games, but on games of 1 hour/mate >>>AT LEAST, or based on a certain minimum depth reached. It could also add its own >>>reply in the book. >>> >>> >>>Serge Desmarais >> >>I think that Ed's EOC concept looks better (where the book is a mere guidance >>for the engine, but needs to be confirmed by calculation). If the engine gets >>more "sceptical" about the book the fewer reference game exist for a given move, >>this seems to be a good way to avoid extreme evals after leaving book (maybe >>still playing hand-selected sharp lines, gambits etc.). >> >>Moritz > > >Looks interesting to me! I don't have Rebel's program, except for the Decade >version. So, you are telling me thet right on the first move, Rebel calculates? > > >Serge Desmarais I do this in Crafty. But it is dangerous. Just do "book random 0" and crafty will do a search on the set of known book moves to choose one. But doing this almost guarantees that it will accept any gambit, because the search is not as deep as a normal search to save time. And accepting some gambits is quite bad. Trying to hold the gambit pawn is generally worse...
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.