Author: Bob Durrett
Date: 14:51:37 10/24/02
Go up one level in this thread
On October 24, 2002 at 16:54:28, Janosch Zwerensky wrote: >>That is even if it was possible for a >>machine to play absolutely perfect chess (i.e. 32-man tablebases!) a human GM >>could still draw with the White pieces by refusing to undertake anything even >>remotely resembling playing for a win by employing a supercautious approach. I agree. But beyond that, there is the aesthetic aspect too. Most people would prefer to watch a hotly contested battle than a Ho Hum GM draw. That's why I prefer to see "High Level Speculative Chess." In that type of chess one or both players takes calculated risks often, as a matter of habit. In the supercautious approach, a ten game match might be tied with ten draws. In the speculative approach, the same ten game match might still be tied but there may be no draws at all. Bob D. > >I think this would come down to the playing style of the perfect player. I'd >expect the drawing path for chess to be narrow enough to allow a very aggressive >perfect player to win 10-0 with black against even the best humans. I do however >not have much reasoning to offer in favor of this belief, I'd have to concede >(though humans have clearly over-estimated drawing chances at least in a couple >of endgames before they were solved by retroanalysis). >On a sidenote, I wonder how high a top GM who has experience in advanced chess >would win over an "unarmed" world champion if he were allowed to consult a >top-end computer during the game. It wouldn't be 10-0, but I guess they would be >ahead clearly. > >>Anyway, because of the above there's never going to be a 10-0 or 20-0 defeat of >>a human world champion especially since 32-man tablebase are impossible (it's >>hard to see how some 10 to the 42nd power number of positions could ever be >>stored and recalled). > >I don't think we can *really* prove that OTB tournament time control perfect >play isn't doable with the right programming on currently available standard PC >hardware (no, I don't believe this of course, but I do doubt that you'd need to >either analyze or store *all* of these 10^42 positions to achieve perfect play). > >Regards, >Janosch
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.