Author: Robert Hyatt
Date: 18:13:19 01/19/00
Go up one level in this thread
On January 19, 2000 at 18:55:42, Dann Corbit wrote: >On January 19, 2000 at 18:34:52, blass uri wrote: >[snip] >>The question is if the move of deeper blue was the right move. >>It is not clear that 36.axb5 was the right move. >So the argument against 36.axb5 was that it is not such a good move? >Every program has bugs. There are a very large number of tunable parameters >with the deep blue machine. Perhaps one (or many) of them was not optimal. > This is totally hopeless. It plays moves too good. It plays moves too bad. Of course Kasparov did the same thing (h6 in round 6 is one that comes to mind, resigning in game 2 is another.) But it seems that humans can do this, but a computer can't. DB has to play the move we think it should have played, or else something is 'suspicious'... twisted logic... >>If it is not the right move then the expression solve for finding axb5 is not >>the right expression. >> >>It is possible that it found this move because of a bug because I saw no tree >>that prove to computers that axb5 is the right move. >There are lots of moves that computers cannot verify. They cannot verify the >NOLOT positions, for instance. If we saw the output of Deep Blue on the NOLOT >positions and did not have SuperGM analysis already available, a lot of us would >think the analysis was wrong or there might be a bug. > >>The case with the nolot positions is different because I think that there is a >>tree to prove the solutions(it is easy for programs to see that playing moves >>not in the tree is bad). >I don't believe in any tree to prove solutions unless it leads to checkmate. A >deeper analysis can always reveal a better move. Indeed, the analysis for the >"... Goes Deep" articles in the computer chess journal shows that there is about >a 17% chance of improvement for each new ply. That means that the odds that we >keep the current selection is (1-0.17) = 0.83. So after two plies, the odds are >.83*.83 = 0.6889, after 3 plies it is 0.571787. After ten more plies it is >0.1551604118721 or only about 16% chance that we keep the same move. > >A five year old looks at a chess board. He sees that he can take the pawn. His >big brother set up a trap to take the bishop. But his dad is watching and sees >a forced mate and kibitzes (he feels sorry for the 5 year old). > >As we move up the ladder, the more deeply we see, the better our choices. I >don't think it will ever stop getting better until you can see all the way to >checkmate. > >In other words, all analysis is tenative. Unless it leads directly to forced >checkmate.
This page took 0.01 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.