Author: blass uri
Date: 13:21:35 02/06/99
Go up one level in this thread
On February 06, 1999 at 14:49:11, Don Dailey wrote: >On February 06, 1999 at 12:48:09, Paulo Soares wrote: > >>Could a chess player with rating FIDE below of the 2000 >>points ELO, win Kasparov with the aid of a program line top >>running on a PC(for example, a PII-450 with 120Mb RAM), in one >>match 2/40? >> I believe that yes, it will be that I am being delirious? >> Already had one match in these circumstances? >> >>Best regards, >> >>Paulo Soares, from Brazil. > >A chess player with a FIDE below 2000 can beat Kasparov without >any help at all. The question is not IF he can beat Kasparov, >but HOW OFTEN. The whole rating system is based on probabilites >and it's possible for a player of any rating (yes, even a raw >beginner who just learned the rules) to beat any player (even >Kasparov on his best day!) Of course the probability in this >extreme example is incredibly low, so low in fact we can say >for all PRACTICAL purposes that it's impossible. > >Suppose you play a random legal move? There is a probability >around 1/20 or so that you will pick A BEST move. I'm assuming >that in many positions there are more than 1 best move, but the >branching factor is approximately 40. For Kasparov >in any given position, there is a probability of perhaps .95 >that he will play a best move. Since Kasparov probably plays >far from perfect chess I'm guessing at .95 figure. The odds >that you will RANDOMLY put together enough best moves to play >a perfect game are extremely low. But it's a lot more complicated >than that. You don't have to put together a perfect game (since >he probably won't either.) Also, it may be good enough to >occasionally pick a good or very good move without it necessarily >being the "best" move. > >We don't have to guess however. Given 2 players with 2 ratings, >we can calculate the winning expectancy of each player. You >will find that a 1900 player has a tiny but non-zero chance of >beating Kasparov. > >In almost every tournament I have ever attended, there will be some >wild upset, crossing several hundred rating points. There is HUGE >difference between 1900 and 2800 though, the odds are extreme in this >case. > >Probably the question you meant to ask is if there is a "reasonable" >chance you (or a 2000 player) could beat Kasparov with help from >a computer. My personal opinion is that your chances are probably >lower than letting the computer play the game! I do not think the chances are lower than the computer. If I had to play with the help of the computer then I believe that 90% of my moves will be identical to the computer. There are cases when I see that the computer does an obvious mistakes so I will play a different move. I have played >games of consultation and non-consultation chess with other players >in the past and it's not always clear if the team is better than >the best player. I think it has a lot to do with the personality >and dynamics of the team. If the weaker player has the strongest >personality, the game will tend to go that players way. I think that if you do right preperations before the match the team is not going to be weaker than one of the players. In this case it is easy to prepare before the match and see if the team human+computer can beat the computer. I believe if the human is intelligent enough to know that he is weaker then the team should be better than the computer without help because there are positions when computers do stupid mistake(for example they cannot identify fortress positions). 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.