Author: Uri Blass
Date: 00:18:01 06/28/01
Go up one level in this thread
On June 27, 2001 at 21:48:44, Robert Hyatt wrote: >On June 27, 2001 at 19:00:09, Uri Blass wrote: > >>On June 27, 2001 at 17:24:25, Robert Hyatt wrote: >> >>>On June 27, 2001 at 11:56:16, Uri Blass wrote: >>> >>>>On June 27, 2001 at 10:42:07, Robert Hyatt wrote: >>>> >>>>>On June 27, 2001 at 04:15:23, Martin Schubert wrote: >>>>> >>>>>>On June 26, 2001 at 12:05:51, Robert Hyatt wrote: >>>>>> >>>>>>>On June 26, 2001 at 09:31:01, Chris Carson wrote: >>>>>>> >>>>>>>>On June 26, 2001 at 09:05:15, Chessfun wrote: >>>>>>>> >>>>>>>>>On June 26, 2001 at 08:53:13, Chris Carson wrote: >>>>>>>>> >>>>>>>>>>Wow, this is some "2100" player. I wish I could do this. I do not see why the >>>>>>>>>>players do not just play the trojan horse, stonewall, KIA, KID, french, dutch, >>>>>>>>>>2.Na3, or better yet, just set up some closed lost position that any "2100" >>>>>>>>>>player could handle and trounce the program. Since the programs are only "2100" >>>>>>>>>>positionally/strategically and not "2400", it should be a piece of cake on a >>>>>>>>>>866Mhz PIII. This must all be a "hoax", these players must be "paid" by the >>>>>>>>>>evil chess empire "Chessbase". This can't be real, close your eyes, cover your >>>>>>>>>>ears, and for crying out loud, the "Emperior" has wonderful clothes on. >>>>>>>>>> >>>>>>>>>>Quick e-mail everyone the memo: Programs are only "2100" >>>>>>>>>>positionaly/strategically/in closed positions. The word has not gotten out yet. >>>>>>>>>>Why hasn't CNN put the word out, it must be a "Chessbase" plot driven by greedy >>>>>>>>>>corporations. >>>>>>>>> >>>>>>>>>Where are you coming up with this 2100? >>>>>>>>>Is it from a post of Bob's where he wrote in certain positions they >>>>>>>>>are 2100?. Please point me to whatever the post was. >>>>>>>>> >>>>>>>>>Sarah. >>>>>>>> >>>>>>>>Yes it is and a couple of others. I did not save the post, I should have, but I >>>>>>>>do not keep a file of posts on people, well one or two posts from time to time. >>>>>>>>:) >>>>>>>> >>>>>>>>I will get off the "2100" bandwagon, I have made my point, unless it is brought >>>>>>>>up again and then I will defend my point of view, but I will keep the offending >>>>>>>>post as proof. Usually I am satisfied if the person just retracts the >>>>>>>>statement, I have to do this myself from time to time. :) >>>>>>>> >>>>>>>>I may look it up in the archives or maybe not. But the "2100" statement was >>>>>>>>made and then defended as the only possible result in "this type of position". >>>>>>>> >>>>>>>>Mark Young and Uri were also posting on that thread and to that message. They >>>>>>>>may have it or can link quickly to it. >>>>>>>> >>>>>>>>Best Regards, >>>>>>>>Chris Carson >>>>>>> >>>>>>> >>>>>>>You can pick any number you want. I will look for a position and move that >>>>>>>even a 1600 player wouldn't play. They happen in comp games all the time, if >>>>>>>you look at enough games. >>>>>>> >>>>>>>2100 isn't magic. In my case, I just used 2100 to mean "bad" and 2600 >>>>>>>to mean "good". >>>>>> >>>>>>"Bad" mistake. If you say "2100" then everyone will remember you that you said >>>>>>"2100". Not "2150", not "1600". Everyone asks you why "2100". Nobody thinks >>>>>>about that you could have meant something different than just the number. What a >>>>>>pity. >>>>> >>>>> >>>>>For those that assume ratings are "absolute", they will probably _never_ under- >>>>>stand anything I write with respect to Elo numbers. I don't believe that I >>>>>_only_ said 2100. My current feeling is that computers play positionally like >>>>>a 2100 player on average. They play some 1600 moves, and they play some 2600 >>>>>moves. So an "average". In some types of positions, they play tactics like a >>>>>2700+ player (these are positions that rarely come up in human vs human games, >>>>>where pieces are pinned, hanging everywhere, lots of threats, etc...) In some >>>>>types of positions they play tactics like a 2100 player. (No I don't think >>>>>they will _ever_ play tactically like a 1600 player, but computers can and do >>>>>fall to very deep/forcing tactical lines simply because the primitive search >>>>>extensions we use don't help everywhere.) >>>>> >>>>>Machines _certainly_ don't play tactically like 2900 players, otherwise they >>>>>would _never_ lose a blitz game. Those that think "blitz is solved as far as >>>>>computers go" just haven't seen the right players... >>>> >>>>If they do not play better than 2600 in nothing they could not get performance >>>>of more than 2700(2600 can be lucky and get performance of more than 2700 in one >>>>tournament but a player that is 2100 in part of the positions and 2600 in part >>>>of the positions has no practical chance to get performance of more than 2700). >>> >>> >>>That is statistically incorrect. A pure 2500 player can be expected to produce >>>a 2700 result in one of every 4 events he plays in. Just because of the >>>statistical anomalies... >> >>No >> >>By this logic a pure 2400 player can get GM norm in one of every 4 events when >>GM norm is possible because this player needs performance of 2600 for GM norm >>and I do not see it happens. > > >If a tree falls in the forest and nobody is around, does it make any noise? > >First, statistics predicts that this will happen. If it doesn't, then the Elo >system is based on faulty math. It doesn't say it will happen one out of every >four events. It says it _could_ average one of every four. There is an equal >probability that the player will perform at a 2200 level also. My original one >of every 4 was not quite correct. For any single game, a player would have a >one in four chance of beating a 2600 player, and giving himself a 3000 TPR for >that one game. If he draws the next 3, he has 3*2400+3000 / 4 for his TPR over >4 games. That is close to 2600. two good games, plus some draws and a couple >of wins over 2400 players and he is there. > >Just because it hasn't been seen, doesn't mean it won't happen. I simply >happen to believe that computers are more likely to see this happen as they >are more consistent and humans tend to overlook them for the moment. Some >are noticing them of course. And many others are actively working to exclude >computers from the equation altogether... > > >> >>There are tactical positions when the machine has no chance against 2600 but >>there are more tactical positions when the 2600 humans blunder. >> >>Uri > > >Sure... but I don't believe it is legitimate to say that a computer is 2900 at >tactics based on that... A 2900 tactical player would be very difficult to >handle. I think that the result is the important thing. Computers get often performance of more than 2600 inspite of the fact that part of the opponents use anti-computer strategy. It is clear that even without tablebases computers are better than 2600 if humans do not try anti-computer strategies.(Kramnik used anti-computer strategy against Deep Junior and there are more examples). It is clear that in order to be better than 2600 in normal games(not trying anti-computer strategy) you must be better than 2600 in something. If you think that it is not tactics then what is it? 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.