Author: Dann Corbit
Date: 11:43:36 01/26/02
Go up one level in this thread
On January 25, 2002 at 20:42:29, J. Wesley Cleveland wrote: >On January 25, 2002 at 17:53:39, Dann Corbit wrote: > >>On January 25, 2002 at 17:49:13, Albert Silver wrote: >>[snip] >>>I disagree. Two things: Heinz's study showed that adding plies doesn't linearly >>>add strength. Second, I think the comparison with Kasparov is amiss. Kasparov >>>does far more than calculate plies, and he would bring that with him in any >>>game. You can take a dry middlegame with no magic ruptures or sacrfices, and >>>Kasparov might tell you in a second that it is a draw. Why? Not because he >>>calculated it to the last ply, but because his judgement and vision allow him to >>>make that assessment. I do not believe for one second that perfect play would >>>suddenly change that. The perfect player might know that h4 and an enormous >>>number of useless moves can or will lead to a loss, but that doesn't mean >>>Kasparov will play them. >> >>I think it's hero worship. If you take a 2400 player against Kasparov, and the >>2400 player is going to get slaughtered for the very reasons that you mention. >>If you take a 3200 player against Kasparov, Kasparov will look just as bad as >>the 2400 player did. Deep Blue, the second version, made Kasparov look almost >>human. A computer that searched 500 times deeper would humble Kasparov. > ^^^ >You mean a program that searches about 7500 ply deep ? Sure. I didn't say it could be done. > >> I believe it would win 1000 out of 1000 games with no draws. > >Unfortunately, the ELO system breaks down for 100% wins. Let's take it in >smaller steps. Assume an ordinary superhuman player, 3500 ELO - Kasparov could >expect to get 1 draw every 50 games. Do you think that this player would also >have no chance against a perfect player ? If so, assume a top superhuman player, >ELO 4200. The ordinary superhuman player could expect to get 1 draw every 50 >games against her. Do you think that *this* player would also have no chance >against a perfect player ? Since ELO is an exponential scale, I really don't know what to think about a 4200 player. I suspect if a perfect player can see clear to the end of the game, we can invent a less perfect player that sees all the way to the end minus one ply (by whatever magic formula). In this case, the perfect player is still better, but the imperfect player will still do pretty well by guessing the last ply based on what they see so far. So I guess my answer depends on the real difference between the two players. If a perfect player has infinite ELO, the opponent cannot score against them if imperfect (finite ELO). But maybe an imperfect player can also have infinite ELO. Or maybe infinite ELO is impossible due to the nature of the game itself. These are interesting questions. Unfortunately, I lack (forever) the special tools to know the answers, and Gedankenexperiments don't seem to provide trustworthy answers either.
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.