Author: Uri Blass
Date: 01:33:09 09/07/02
Go up one level in this thread
On September 07, 2002 at 01:10:12, Dann Corbit wrote: >One point to consider is that 90% correct move choices means certain failure >within only a few plies. > >First ply :9/10 chance that the program makes the best move. > >Second ply will only be correct IF the first ply was also correct. If the first >move chosen was bad, then this is still bad. So we multiply the probabilities >since the events are dependent. Hence 9*9/(10*10) = .81 chance > >3rd: .729 chance we are OK so far, if the program guesses right 90% of the time > >4th: 0.6561 > >5th: 0.59049 > >6th: 0.531441 > >7th: 0.4782969 > >8th: 0.43046721 > >9th: 0.387420489 > >10th: 0.3486784401 ... by now the chance is about 1/3 that we have not made a >blunder with an engine that guesses right 90% of the time. By move 20 the odds >are 0.12157665459056928801 and by move 30 we are down to 4% chance. > >So it is clear that a program that almost always guesses right is almost certain >to fail very early in the game! Dependent of the length of the book line. If the book line is 60 plies you are right. Inspite of this I do not consider opening knowledge as the most important knowledge at this moment because even after a move that is not the best the opponent needs to find the correct reply so practically in most games against equal opponents book is not the main problem of movei and today I can earn more rating by improvement of the endgame knowledge. Note that 90% was only an estimate. I did not calculate it and it is also hard to calculate it because not always we know the right move. Uri
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.