Author: Dann Corbit
Date: 14:38:07 07/12/01
Go up one level in this thread
On July 12, 2001 at 17:19:29, Uri Blass wrote: >On July 12, 2001 at 16:59:45, Dann Corbit wrote: > >>On July 12, 2001 at 15:35:47, Larry Oliver wrote: >> >>> >>>Starting with its book turned off and left to its own devices, could todays top >>>programs on a very fast computer with say 12 hours per move think time, invent >>>the standard openings? >> >>Have not tested it lately, but Crafty used to like to play the Sicilian dragon, >>if you remove the opening book. > >Larry suggested to test it at 12 hours per move on a very fast computer(see >above). This will add one full move to my suggested 7 full moves (so 16 plies) at most [unless you do risky extensions -- and then what?]. Even one week will not help [18 plies -- not quite 19]. One month will probably produce useful results, but who do you know who is willing to wait so long? I think that much better results will be had by analysis of forward positions. Along those lines, there are 7 million online games between rated players which someone has in a collection, but I can't get at. Anyway, add that to the 3 million games I have collected and you will have about 10 million games (give or take a million). Now, with this data, analyze every move that has ever been played (less than one billion). Minimax the result of that analysis and you will surely be able to produce some good results. I have done experiments with a powerful computer thinking for 24 hours. Sometimes it finds something brilliant, and sometimes it says something stupid. 24 hours is definitely not enough, not even with a fast machine and not even with a good program. If you want to make better opening books, that is. [snip]
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.