Author: Peter Berger
Date: 06:53:49 03/24/03
Go up one level in this thread
On March 24, 2003 at 09:42:49, Drexel,Michael wrote: >On March 24, 2003 at 03:56:46, Uri Blass wrote: > >>On March 23, 2003 at 20:03:15, Peter Berger wrote: >> >>>On March 23, 2003 at 12:06:23, Uri Blass wrote: >>> >>>>On March 23, 2003 at 11:51:52, Peter Berger wrote: >>>> >>>>>On March 23, 2003 at 04:17:22, Frank Phillips wrote: >>>>> >>>>>>On March 23, 2003 at 02:38:31, Uri Blass wrote: >>>>>> >>>>>>>On March 22, 2003 at 23:15:19, Lyn Harper wrote: >>>>>>> >>>>>>>>On March 22, 2003 at 13:17:48, Uri Blass wrote: >>>>>>>> >>>>>>>>>On March 22, 2003 at 13:16:46, Uri Blass wrote: >>>>>>>>><snipped> >>>>>>>>>>The only reason to make them weaker relative to humans is simply to change the >>>>>>>>>>rules of the game. >>>>>>>>> >>>>>>>>>Should be the only way to nake them... >>>>>>>>> >>>>>>>>>Uri >>>>>>> >>>>>>>I see that I made a mistake in my correction (make and not nake) >>>>>>> >>>>>>>> >>>>>>>> But is'nt this just inventing ways to delay the inevitable? The programs are >>>>>>>>just getting stronger while the humans are'nt. Accept it. >>>>>>> >>>>>>>If you change the rules humans will be relatively stronger and after the delay >>>>>>>you can change the rules again. >>>>>>> >>>>>>>I think that humans also can learn to be stronger in normal chess thanks to >>>>>>>computers. >>>>>>> >>>>>>>I agree that in every static game computers are going to win after enough time >>>>>>>and this is exactly the reason to change the rules. >>>>>>> >>>>>>>Uri >>>>>> >>>>>> >>>>>>So we have a game the rules of which are that humans must be able to >>>>>>win....bizarre. >>>>>> >>>>>> >>>>>>Frank >>>>> >>>>>Why is that bizarre? It's just a little early now but some day some kind of >>>>>rules will be needed to make man-machine matches interesting as the silicon will >>>>>be too strong for equal competition. >>>>> >>>>>Obviously there are two ways this can be done: >>>>> >>>>>a.) limitting the computer power and ressources >>>>> >>>>>I like this approach most. Will there be a day when a program on a current >>>>>Pocket PC or Palm can compete with human top players? That's a very long way to >>>>>go still. >>>>> >>>>>b.) adapting the rules >>>>> >>>>>Maybe the easy and old approach of playing with uneven material is nicest here. >>>>>That's a little similar to Go. How long until a program can win against top >>>>>players with a knight less? >>>> >>>>Never >>>> >>>>I believe that even god is going to lose against the top humans >>>>and even against 2600 GM's with a knight odd. >>>> >>>>Uri >>> >>>I think you underestimate God - he would have a few additional advantages, for >>>example he could read his opponents' minds. >>> >>>Seriously - I don't think knight's odds is something that can't be overcome with >>>extremely deep searches at some time. Or if it is, make it one or two pawns >>>instead. >>> >>>This year Mr Ingo Althoefer did a few experiments playing chessprogrammers and a >>>few strong players in odds games with the help of an engine. He seemed to do OK. >>>Maybe like 2100 level. >>> >>>That's were chessengines were 25 years ago, too. I don't see the principle >>>difference. >>> >>>Let's assume we played a corresponcence game and I gave you knights odds but >>>earned the right to use a chessengine where you could only use your own mind. I >>>would be pretty confident to win or at least I bet you would have a very hard >>>time. >> >>I believe that I can win a game in these conditions. >>I do not think that it going to be easy but I am not a GM. >> >>I also think that knight is equal more elo at the high level and even at 2000 >>level at 120/40 it equals many hundreds of elo. >> >>Computer may have chances against GM's with knight odd only in blitz. >> >>Uri > >No way, >I played three 5 min test blitz games against the top programs Fritz 8.008 and >Shredder 7.04 (both playing without Nb1) on Athlon 2200+ with the black pieces >and won three times in a row. >My level is 2300 and I am really not especially strong in blitz. >I claim to win at least 9:1 against any top program in a ten games match with >these conditions. >A Grandmaster should do even better. >To win a correspondence game with a piece up is of course no problem at all. > >Michael > This discussion is of course a little theoretical. It assumes a.) Some more time has passed and computers have become even stronger due to hardware and software advances. b.) Programs are tuned for odds games. Currently it is not so difficult to win against them because they are not programmed to do well in this setup. I think a program together with a clever operator (maybe 1800-2000) who has the power to overrule might be at 2100 level in knights odds games currently - of course I have no data to present but it should not be too far off. The clever operator can be replaced with clever algorithms later. This is the same situation as with regular chess 20 years ago. I might be wrong but then make it two pawns odds - the actual material difference is not so interesting. I also disagree when it is about correspondence compaired to blitz - I think it is easier for humans to win odds games in the described setup in blitz now.
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.