Author: Robert Hyatt
Date: 13:46:29 11/15/00
Go up one level in this thread
On November 15, 2000 at 14:02:57, Alvaro Rodriguez wrote: >On November 15, 2000 at 13:06:39, Robert Hyatt wrote: > >>On November 15, 2000 at 12:49:20, Alvaro Rodriguez wrote: >> >>>On November 15, 2000 at 11:04:04, Robert Hyatt wrote: >>> >>>>On November 15, 2000 at 00:55:51, Alvaro Rodriguez wrote: >>>> >>>>>On November 14, 2000 at 17:30:44, Robert Hyatt wrote: >>>>> >>>>>>On November 14, 2000 at 14:14:22, Alvaro Rodriguez wrote: >>>>>> >>>>>>>On November 14, 2000 at 13:09:55, Mogens Larsen wrote: >>>>>>> >>>>>>>>On November 14, 2000 at 13:00:53, Alvaro Rodriguez wrote: >>>>>>>> >>>>>>>>>It´s a big deal because the program expects to play a human opponent and when >>>>>>>>>the person who is running the (C)(in this case Rebel Tiger) analyzes the games >>>>>>>>>played against humans, he will analyze this game too and import it to a database >>>>>>>>>with all human opponents. So, he will get wrong results.. Allthough, it´s just >>>>>>>>>one game but if everybody cheats, then the results against humans will be >>>>>>>>>completly worthless. IMO >>>>>>>> >>>>>>>>A good point if it's important to distinguish between human and computer >>>>>>>>opponents when analysing your games. However, a lost game is a lost game and >>>>>>>>should analysed due to that fact alone independent of the opposition. >>>>>>>> >>>>>>>>I'm still of the opinion that it's rather harmless to play "advanced" chess >>>>>>>>against a computer opponent compared to cheating against a human player. But >>>>>>>>maybe I'm just against the idea of computer program "rights" in general :o). >>>>>>>> >>>>>>>>Mogens. >>>>>>> >>>>>>>Yes, if I had my own program to operate, I would not distinguish between humans >>>>>>>and computers when analyzing the games, but I guess there is a difference. Bob >>>>>>>for example, would care as he prefers to play against the GM´s on ICC and >>>>>>>analyzes those games with more interest then against computers. It depends on >>>>>>>who is operating, but I guess most of the operators cares wheather it´s a human >>>>>>>or a computer. >>>>>> >>>>>>In general, on ICC, (with respect to book learning) I don't care about >>>>>>silicon vs carbon. I pay attention the _rating_. And a cheater can sure >>>>>>wreck this... ie he is rated 2500 but chooses to not use the computer and >>>>>>plays like an 1800 as a result. Or he is 1800 but uses a computer and plays >>>>>>like a 2500. If I take the ICC rating and factor that into the learning >>>>>>formula, I get wrong answers. >>>>>> >>>>>>I also want to know I am playing a computer as I am more cautious about setting >>>>>>the contempt factor. Against humans, the contempt is a function of their rating >>>>>>and Crafty's current ICC rating. Against computers, contempt=0, period, as >>>>>>doing anything else lets the opponent influence the game by fiddling with the >>>>>>contempt. >>>>> >>>>>This is a reason why silicon vs carbon matters. If you set the contempt=0 >>>>>believing you´re going to play a computer, but instead, the operator plays >>>>>himself. Then, you get wrong results. If you play humans only on ICC, do your >>>>>rating get higher then if you play only computers ? >>>> >>>> >>>>Depends on what you mean by "you". >>> >>>Sorry, I wasn´t clear, I meant crafty. >>> >>>If you mean me as a human, my rating >>>>would probably be higher if I only played computers. How much is a guess, >>>>of course. But I would _definitely_ play things I would not play against >>>>humans. >>> >>>Really? Your rating would be higher playing against computers? It´s the opposite >>>for me, I find it easier to play against strong humans then strong computers. >>>Therefore, my results are better. >> >> >>There are some chess engines around that play like 2600 players if you play >>open, traditional chess. They can also play like an 1800 player if you stuff >>the board and don't give them any room to launch some wild tactics. I can >>show you many players on ICC that produce excessive numbers of draws against >>programs. And since the programs are always rated much higher than the human, >>the human's rating is dragged upward by doing this. >> >>I can't beat programs very often, but I can certainly beat them often enough >>to convince me that they aren't GM players. I doubt I would _ever_ beat a GM >>player, using the same strategy I would use to try to draw a chess program. >> > >This is true. You can´t use the same strategy against comps as you use against >humans. You can´t play anti-computer strategy against a human, well, I guess you >can but I bet it doesn´t work out very well. I´ve played some GM´s in >Worldchessnetwork, occasionally won some games, in 5 min time controls. But I >don´t think I can win over programs in 5 min games, but maybe in a long game, I >have a bigger chance against a comp then against a human. > I generally believe that my results against humans are better when I play the kind of tactical game I like to play. I would _never_ play a strategic game against a GM. I used to play IM Mike Valvo a lot at ACM events (he was always playing everybody at 5:1 time odds). I always played wild openings against him, because I felt that I might out-calculate him here and there, but I would _never_ "out-experience" him anywhere. >> >> >>> >>>If you mean "crafty" then its rating will be higher playing humans >>>>rather than computers, for lots of reasons. Tuning. Computer operators are >>>>flakey at times, running a weak program, then a strong one, or a slow machine >>>>followed by a fast one. Such computer opponents have a rating that is never >>>>matched to the machine/program that is actually playing at the moment. If >>>>you catch him on weak hardware with a high rating, you win points. If you >>>>catch him on fast hardware with a low rating, you lose points. Sort of a >>>>crap-shoot in many cases. >>> >>>Yes, you have to consider this also, that the operator can switch programs from >>>time to time. But if crafty plays 100 games against humans rated 2300 and 100 >>>games against computers rated 2300(without switching from the "original" program >>>to another program who is NOT rated 2300) , does crafty score better against the >>>humans ? (AC) accounts should be taken into consideration by the ICC directors. >>> >>>Alvaro >> >> >>I personally think crafty would do better against humans. It has some code >>to make it extremely difficult to block the position, except for those cases >>where the position gets blocked before it gets out of book. This code tends to >>make it accept 'concessions' to avoid the blocked positions, concessions that >>might not be a good idea vs a computer. > >I think a bad opening for programs in general, I have no particular games to >show this, is 1. e4 d6 2. d4 Nf6 3. Nc3 g6 etc.. Black often gets a chance to >play f5 followed by f4, closing up the position and preparing for a kingside >attack which gives computers a hard time for the rest of the game. I´ve seen >some games with this theme for sure.. Does crafty avoid this position, or does >it let black play f5 followed by f4 ? > >Alvaro >> >> >> >> >>> >>>> >>>>I wish this didn't happen, and in the case of _my_ program, the hardware is >>>>a constant. As is the program (excepting when I make changes of course.) >>>>Others don't do the same, however. Which makes the rating pool fluctuate more >>>>than would normally be expected. >>>> >>>> >>>> >>>> >>>>> >>>>>Alvaro >>>>>> >>>>>> >>>>>> >>>>>>> >>>>>>> >>>>>>>Advanced chess is interesting for sure, but I think you should tell your >>>>>>>opponent before the game if you are about to use a program as a "coach". >>>>>>>Otherwise, your opponent thinks he´s playing against a human and that is wrong >>>>>>>IMO. Mogens, what do you mean when you say you are against computer program >>>>>>>"rights" in general ? >>>>>>> >>>>>>>Alvaro
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.