Author: Dann Corbit
Date: 21:06:16 06/15/99
Go up one level in this thread
On June 15, 1999 at 23:52:43, Laurence Chen wrote: >On June 15, 1999 at 23:29:16, Dann Corbit wrote: > >>If you have an opening book that contains brilliant moves -- especially >>positional or sacrifice based upon completion -- it is quite likely that they >>will cause your program terrible harm. Having a brilliant move is of no >>benefit, if your program does not know what to do with the position. Even if >>the opening book suggests the next move, unless your program can see what to do >>after that, having such a position could do a lot more harm than good. Being >>able to utilize such a position means that you must exploit a plan that >>understands the position. >> >>Opinions? >If you want to be completely scientifically about, it is best to run two control >tests. One with a prepare opening book with brilliant moves, and another one >without it. Control test 1 will use the brilliant modified book, and play >against engines without this new modified book. Control test 2 will use the same >opening books without brilliant moves used by all the engines. This will tell >you if such brilliant moves are harmful or not. An excellent suggestion. I think that the hard part here is finding the brilliant moves that the computer does not understand. I suspect it will be easier to simply take a list of perhaps 100 moves from test suites which are known to have a bm that the computer does not find. Create two test sets: 0. Use the bm and ignore the computer suggestion 1. Igore the bm and use the computer suggestion Then from those positions and with those conditions, let two computers complete a game. Reverse the computers and positions and repeat. Then repeat for each data point in the set. With something as tricky as this, I think it would take a lot of trials to be sure. I thought that perhaps some programmers who are intimately familiar with their programs would know whether such moves would help or harm by a gedanken experiment.
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.