Author: Sune Fischer
Date: 12:29:29 03/20/02
Go up one level in this thread
On March 20, 2002 at 14:54:22, Dann Corbit wrote: >On March 20, 2002 at 14:07:18, Sune Fischer wrote: > >>On March 20, 2002 at 13:58:35, Gian-Carlo Pascutto wrote: >> >>>On March 20, 2002 at 13:50:47, Sune Fischer wrote: >>> >>>>It is not the limit, chess is finite so you can never reach an infinite elo >>>>because you simply can't win them all. >>> >>>This was refuted earlier in the thread, i.e. in the case >>>of deterministic players. (like chessprograms!) >> >>Not really, see there was a trap built into it :) >>They are in fact playing the _same_ game over and over, so it should only counts >>as one I think :) > >If you could beat someone with a simple "fool's mate" style kingside attack each >and every time, would it count as only one win, or a win for every time you >executed the attack?[1] No matter how weak, they sometimes beat you! I've seen a 1300 player beat a 2150 player, and a 1600 beat a 2400, okay it was blitz but still there is always a chance :) It would be great if we could pull some statistics from the servers, I would like to know how often this actually happens. >>Anyway, thought about the limited set, and I figure, that if A beats B 2 out of >>2 games, and B has a rating of 1500, then we should give A a rating that would >>give A a 50% chance of scoring 2 of 2 against B, this is all we can really say >>since we have no idea of how the third game will end. This means A should have >>probability 0.7071 of a win. I don't have the formula in front of me, but it's >>probably something like 1650. > >You have to have a very large number of games to make any sort of meaningful >predictions. You can (of course) have any ELO figure you want for either/both >players in this simple system. For instance, if the lower ranked player is 100 >ELO below the stronger one, and you want the stronger one's rating to be 'x', >then just set the weaker player to x-100. It's completely arbitrary. Only the >differences matter. Sure, but still finite mind you :) I get the feeling that some people do not fully understand the meaning of _infinite_ ;) >>But hey, I'm not a mathematician, so I could be wrong :) Come to think of it, this is a maximum likelihood sort of problem, finding the most likely parameter is basic statistical analysis. >[1] I like to use this technique to teach little kids how to play chess. Beat >them with it over and over, explaining the reason exactly why I am moving each >piece. After a dozen games or so, they will be able to defend because they will >figure out how to stop it.[2] The side benefit is that they can defeat their 6 >year old friends quite a few times before they catch on and it makes them feel >pretty smart. Of course, it is a truly awful opening, and I have problably >damaged them for life by teaching it to them[3] > >[2] Usually, they figure it out by trying to attack me with it. As I explain >why I am moving the pieces to defend, they get the idea. Once in a while they >figure it out from first principles. > >[3] But think of what happened to Fischer and Morphy. I may have saved them >from a life of bizarre brain dysfunction :) -S.
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.