Author: pete
Date: 13:52:13 06/06/00
Go up one level in this thread
>>>[D]r2qr3/pb3pk1/np3b1N/2pp3Q/3P4/1PR1P1P1/P4PBP/5RK1 w - - 0 1 >> >>What is strange to me is that Dr Hyatt has posted that Crafty17.11 sticks to >>Nxf7 right from the start until end of ply 14 . This is really a major change to >>17.10 and the result seems to be not too good . > >I think this is a good example to show that "best moves" in EPD test positions >are only "best moves" for a certain depth unless a checkmate is found. > >For instance, a computer of ten years ago would be doing very well to find Nxf7. > But upon searching deeper, it seems g4 might be better. After an even deeper >search, e4 looks better. > >I would not be surprised in the least if an even deeper search turned up >something else, including one of the previous moves. > >In short, I think that best moves are valid to some particular ply depth if they >have the appropriate evaluation. But until we definitely find a checkmate, >there is always a possibly better move -- perhaps hidden by NULL move or some >such. I don't agree here though your general idea seems convincing a.) what you say about first Nxf7 , then g4 ; then maybe again Nxf7 , then e4 !at last seems right to me . So a program starting with Nxf7 and never changing seems strange to me . b.) But this position is analyzable ; for example you simply have to play a few moves of the mainlines ; then programs will have an easy time detecting the truth. Often it _is_ possible to find out the truth about chess positions , sure it happens that years later a refutation is found but even before computers were thought off humans were analyzing chess positions and most have passed the test of time . Look at the great chess literature like "New York 1924 " , "Alekhine's best Games" , "Fisher's 60 memorable games" The analysis has survived all the deep analysis of chessprograms ( as for example CAP proves every day ) . From time to time a little glitch is found but the more radical a grandmaster judges the more often he was found to be right even if it seemed to be uncalculable before . Sure , maybe some super computer will once announce mate in 37 after Nxf7 but it just isn't that likely . It hasn't proved right so far that chessbooks have to be written in a new way because computer found zillions of misjudgements of human mind . Even the endgame databases in general haven't found many _really_ new things . In most even complicated positions humans have known the right plans before although it was completely uncalculable to say something like mate in 137. To say every new generation of computer hardware will let a new move look like the best ones neglects that in fact a wide territory of chess is already understandable to humans ( although only to very few :-) ) and chess is no random game . IMHO for sure . pete
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.