|
Main
Date: 26 Apr 2006 17:09:29
From: Ray Gordon
Subject: There is PLENTY of room at the top in chess
|
Look at the top players against the top computers: they can't compete. What this means is that even our "Super GMs" are quite beatable. If Hydra and Fritz can do it, so can a human. We just aren't as good at chess as we like to think. -- "Google maintains the USENET." -- The Honorable R. Barclay Surrick, Eastern District of PA Judge From Parker v. Google, E.D.Pa. #04-cv-3918
|
|
|
Date: 28 Apr 2006 07:56:42
From: Ralf Callenberg
Subject: Re: There is PLENTY of room at the top in chess
|
Ray Gordon schrieb: > >>>> So being booked up perfectly to move 30 is "here and there?" > >>> People are still not perfect to move 30. > >> > >> Maybe not YOUR people.... > > > > Who is "YOUR people"? Those from planet earth? > > Those who aren't booked up to move 30. There are players who are. That people are prepared up to move 30 I don't doubt. But that they are "perfect" is completely exaggerated. > It is > > simply impossible for humans to calculate to the depth as current top > > programs do. Added to this, humans make errors. > > Only when they don't know what they are doing. No. The experience of thousands of years of civilization has shown: to err is human. Man's brain is not built to be perfect. They will always make errors. And if one is handling long and complicated variations it is nearly sure that he will make errors, so it doesn't make sense to go any further. > >They forget about pieces or put pieces on squares where they no longer are > >standing, simply oversee variations. If you calculate longer variations, > >the probability for such errors grows very fast. > > For weaker players, it would. No, for stronger players it just takes longer, but even GMs have their limits. You can not keep endless long variations in memory and still visualize all positions with the same clarity. > I didn't say it was easy to duplicate a > computer's performance at chess, but it is definitely possible to do this or > even to exceed it. No. It is simply impossible. The human brain is not capable of performing such tasks in this perfection. > > At least up to now there is no hint into this direction. On the > > combinatorical side, the computers' advantage is still increasing. > > Over the *average* human. Also over GMs. > >> We don't have to calculate every variation the way they do, but we > >> certainly can improve that area sufficiently to neutralize the computer's > >> *only* advantage over us. > > > > This is just an assumption. There is no hint, that this might be the case. > > Computers don't have humans' "chess vision," especially in the opening. Well, look at Hydra. The programmers strategy was to implement only a comparable small opening library and then let Hydra play alone. It has proven to be quit successfull, just ask Adams. > > Show me a machine that won't play dxc5 if Black leaves it "hanging" in a > queen-pawn game. > So? He can play it, because he is strong enough. > > Those draws had other reasons. They didn't appear because both sides had > > reached perfection. > > The "other reasons" were based on the idea that getting a win for either > player was "extremely difficult" due to the small amount of mistakes. It's > two sides to the same coin. And then, why were there wins and losses on the next match? Did they somehow lose their perfection? > > > >>>> Let's just say that Fischer's 60 memorable games aren't so memorable > >>>> anymore. > >>> Whoever says something like that doesn't know too much about chess. > >> > >> Or knows a lot more.... > > > > Which GM said so about this book? > > They didn't have to. It comes out in the course of analysis of other games > when they note the improvements. The games may be memorable for the fans, > but they are no longer so relevant to chess theory. They are relevant for those who want to study chess. That you don't learn opening theory from them is obvious. Nevertheless I assume, most GMs have read it at some point. > Still, pattern recognition is the heart of chess improvement. If I didn't > know how to improve my chessgame, my rating would never have hit 2000 back > in the 1980s. Sure pattern recoginition is important. But with it alone you simply can not play. You have to sit down and calculate the variations. Pattern recognition is of help, but it does not take all the burdens from you. > > >>> It is an unproven claim that this is possible in a way much beyond what > >>> is possible for humans today. > >> > >> All theory is unproven at first. > > > > Well, but at least there is some evidence in the first place. Could you at > > least show this? > > The priy evidence is that computers aren't spitting out perfect moves or > anything close, and that a lot of the time they have the wrong side winning > the game. OK. > That means it's *possible* for a human to exploit this. No, your logic is flawed. It only shows, that it is theoretically possible to play better than todays computers. This is absolutely no evidence for the claim, that it must be possible for humans to reach the same strength without calculation. > >>> As humans don't play perfectly this is an uninteresting question. > >> > >> Which humans are you talking about? > > > > The living ones. > > The original statement referred to perfect opening play. OK. They are still not perfect. > > You claim you are perfect. Well, I think no other GM has done this before. > > Congratulations. Seriously: as long as you didn't prove this in a real > > tournament (no blitz or bullet crap), against strong players, I simply > > don't believe you. > > I never said I was perfect for the whole game, just that I don't make many > mistakes in the opening. I understood what you were saying. Computers have made that even easier to accomplish > now. OK. Did you ever switch off the opening theory for Fritz and played against him, or one of those new monsters, which you might not know and are not prepared for? > > Well, I can only repeat: play in a tournament and show it. Otherwise it is > > simply a claim, not more. > > Want to pay my entry fee and compensate me for missing two days of work? Of course not. You spent thousands of hours with training and you don't have the time and the few buckt to spend a few days for a tournament? > > The reason I don't play now is that it's a big work in progress, and people > only judge based on what they see now. Well, you said, you have improved 150 to 200 points each year. This makes you a player of GM strength now. This is where you claim to be *NOW*. And this you could prove. > I could come back, get a 2300 performance rating, and people would see that > as 500 points below where I need to go. I could be right on schedule to get > there, but it wouldn't matter. For that privilege, I could waste time and > entry fee, instead of staying home and working on my game. It's your decision. My decision is, that I simply don't believe that you gained this strength. > > Sure, but there are only few concepts compared to what a GM knows. And a > > GM can find a move even without calculation, just based on positional > > knowledge. > > Which is how it's possible to "outcalculate" a computer. Well, this is the hope. > > >Computers are not able to do so. They have to calculate. > > So do humans. But not to this extent. Computers can do *nothing* without calculation. But as they can calculate like hell, this results in monsters on the board. Humans can not calculate so fast and precise as computers, so the have to rely on other factors. That they can improve this so much, that they can stay ahead of the computers, is just a hope and there is not much evidence, that this is the case. > > > >>> Sure.If They can calculate 10, 15 moves ahead, they might see, that this > >>> might be an advantage. No sophisticated knowledge is required. > >> > >> They sure aren't finding checkmate with a move like that. They are > >> evaluating the position at the end of the horizon to accomplish this. > > > > Yes, but a GM could make such a move based on general assumptions. He > > doesn't have to calculate 10 or 15 moves. > > How have the "general assumptions" performed so far for players like Adams > against Hydra? Not impressive. That's exactly what I am saying: the computers' style is completely different from a GM. And it doesn't look like the GMs will keep the edge much longer over those number crunchers. > > >Again: a computer has only very small knowledge about positions. Only if > >you combine this with his ability to calculate a lot of variations, you get > >the strength we can see. If you would take the existing evaluations and > >stop the program thinking beyond move 3, you are back to the 80s with > >programs just good enough for amateurs. > > In the 1980s, Deep Thought was rated 2265. Yes, amateur strength. But I meant more those machines which only calculated 3 moves. Deep Thought calcualted much deeper and was singular at his time. > 1956 is "mediocre?" Well, I don't juged by numbers of people in the classes, so I am not talking about average, but by drawing a line between the top and the beginners. 1900-2000 is somewhere in between. So yes, I regard 1956 as somewhat mediocre (and my 2150 Elo as well). > > I don't get it. First: I didn't say anything about the importance, I just > > said, my rating is not very high. Besides that I don't think, that the > > rating is actually so important in a debate. > > So I have to prove myself, but you can get away with already being weaker > than me yet still with a better grasp of chess theory than me I never claimed to have a better graps of chess theory than you. To the contrary I would assume, that your knowledge is much deeper. I just win some games, because I am a fighter, not because I am rekable in chess theory. > > My rating doesn't tell you whether I am weaker or stronger as you are. > > So you're proud of clouding what is supposed to be an open debate? What open debate? Whether you are stonger than me? I simply don't care. You were about my class when you stopped playing tournaments, now after so much effort you put into chess, I would assume, that you are much stronger than me. How much, I don't know. > My original post merely pointed out that GM play is still weak enough for a > dedicated, talented player to come in and dominate. I understood this. But just the logic you based this point on is flawed in my opinion. Greeetings, Ralf
|
| |
Date: 28 Apr 2006 18:05:58
From: Ray Gordon
Subject: Re: There is PLENTY of room at the top in chess
|
>> >>>> So being booked up perfectly to move 30 is "here and there?" >> >>> People are still not perfect to move 30. >> >> >> >> Maybe not YOUR people.... >> > >> > Who is "YOUR people"? Those from planet earth? >> >> Those who aren't booked up to move 30. There are players who are. > > That people are prepared up to move 30 I don't doubt. But that they are > "perfect" is completely exaggerated. You can't book up unless you do it "perfectly," because then you're not booked up, but just plain unsound. >> It is >> > simply impossible for humans to calculate to the depth as current top >> > programs do. Added to this, humans make errors. >> >> Only when they don't know what they are doing. > > No. The experience of thousands of years of civilization has shown: to > err is human. Man's brain is not built to be perfect. They will always > make errors. And if one is handling long and complicated variations it > is nearly sure that he will make errors, so it doesn't make sense to go > any further. When's the last time you made a mistake tying your shoes? The Mott Street Chicken doesn't make errors in tic-tac-toe, either, and will play anyone to a draw or win. >> >They forget about pieces or put pieces on squares where they no longer >> >are >> >standing, simply oversee variations. If you calculate longer variations, >> >the probability for such errors grows very fast. >> >> For weaker players, it would. > > No, for stronger players it just takes longer, but even GMs have their > limits. You can not keep endless long variations in memory and still > visualize all positions with the same clarity. One need not see "all" positions and moves, just the key ones. >> I didn't say it was easy to duplicate a >> computer's performance at chess, but it is definitely possible to do this >> or >> even to exceed it. > > No. It is simply impossible. The human brain is not capable of > performing such tasks in this perfection. So far, neither is a chess computer. >> > At least up to now there is no hint into this direction. On the >> > combinatorical side, the computers' advantage is still increasing. >> >> Over the *average* human. > > Also over GMs. Over *today's* GMs. >> >> We don't have to calculate every variation the way they do, but we >> >> certainly can improve that area sufficiently to neutralize the >> >> computer's >> >> *only* advantage over us. >> > >> > This is just an assumption. There is no hint, that this might be the >> > case. >> >> Computers don't have humans' "chess vision," especially in the opening. > > Well, look at Hydra. The programmers strategy was to implement only a > comparable small opening library and then let Hydra play alone. It has > proven to be quit successfull, just ask Adams. Adams is not strong in the opening. >> Show me a machine that won't play dxc5 if Black leaves it "hanging" in a >> queen-pawn game. >> > > So? He can play it, because he is strong enough. IT plays it because IT doesn't understand a) that the pawn isn't really hanging; and b) that it's antipositional to do so. >> > Those draws had other reasons. They didn't appear because both sides >> > had >> > reached perfection. >> >> The "other reasons" were based on the idea that getting a win for either >> player was "extremely difficult" due to the small amount of mistakes. >> It's >> two sides to the same coin. > > And then, why were there wins and losses on the next match? Did they > somehow lose their perfection? Karpov broke down physically and mentally. I watched it happen. Had you been born yet when this match took place? >> >>>> Let's just say that Fischer's 60 memorable games aren't so memorable >> >>>> anymore. >> >>> Whoever says something like that doesn't know too much about chess. >> >> >> >> Or knows a lot more.... >> > >> > Which GM said so about this book? >> >> They didn't have to. It comes out in the course of analysis of other >> games >> when they note the improvements. The games may be memorable for the >> fans, >> but they are no longer so relevant to chess theory. > > They are relevant for those who want to study chess. That you don't > learn opening theory from them is obvious. Nevertheless I assume, most > GMs have read it at some point. I began my study of openings by copying Fischer's repertoire and have spent as much time going over his games as almost anyone you'll ever come into contact with. However, after all this study, I have found even his games and style of play to have weaknesses, although to exploit them, one must first methodically acquire his level of technique in order to win a won position or save a draw against him. That is what separates the top players from the near-greats, all else equal in the opening. The only player I have never been confident that I could crack even with perfect technique is Kamsky, and I believe he's our best hope for a world champion, even over Nakamura. I saw Kamsky lose over 20 consecutive games to a silicon monster online one night actually. >> Still, pattern recognition is the heart of chess improvement. If I >> didn't >> know how to improve my chessgame, my rating would never have hit 2000 >> back >> in the 1980s. > > Sure pattern recoginition is important. But with it alone you simply > can not play. You have to sit down and calculate the variations. > Pattern recognition is of help, but it does not take all the burdens > from you. Without knowledge of the patterns (like the smothered mate, for example), calculation becomes much more difficult. >> >>> It is an unproven claim that this is possible in a way much beyond >> >>> what >> >>> is possible for humans today. >> >> >> >> All theory is unproven at first. >> > >> > Well, but at least there is some evidence in the first place. Could you >> > at >> > least show this? >> >> The priy evidence is that computers aren't spitting out perfect moves >> or >> anything close, and that a lot of the time they have the wrong side >> winning >> the game. > > OK. k. > > That means it's *possible* for a human to exploit this. > > No, your logic is flawed. It only shows, that it is theoretically > possible to play better than todays computers. Right: possible. See above. >This is absolutely no > evidence for the claim, that it must be possible for humans to reach > the same strength without calculation. How the human does it isn't relevant. Calculation is obviously part of chess, but that doesn't refute what I say. >> >>> As humans don't play perfectly this is an uninteresting question. >> >> >> >> Which humans are you talking about? >> > >> > The living ones. >> >> The original statement referred to perfect opening play. > > OK. They are still not perfect. Some are. 1. e4 is perfect through move 1. >> > You claim you are perfect. Well, I think no other GM has done this >> > before. >> > Congratulations. Seriously: as long as you didn't prove this in a real >> > tournament (no blitz or bullet crap), against strong players, I simply >> > don't believe you. >> >> I never said I was perfect for the whole game, just that I don't make >> many >> mistakes in the opening. > > I understood what you were saying. > > Computers have made that even easier to accomplish >> now. > > OK. Did you ever switch off the opening theory for Fritz and played > against him, or one of those new monsters, which you might not know and > are not prepared for? I'm prepared for any opponent at any time. When I lose games, it's not because of the opening. >> > Well, I can only repeat: play in a tournament and show it. Otherwise it >> > is >> > simply a claim, not more. >> >> Want to pay my entry fee and compensate me for missing two days of work? > > Of course not. You spent thousands of hours with training and you don't > have the time and the few buckt to spend a few days for a tournament? Why would I bother? I'll play when I'm confident I can win an under-2000 section, a small open tournament, or grab a GM or IM norm. >> The reason I don't play now is that it's a big work in progress, and >> people >> only judge based on what they see now. > > Well, you said, you have improved 150 to 200 points each year. This > makes you a player of GM strength now. This is where you claim to be > *NOW*. And this you could prove. Wrong. I was 1900 (2000 right before then) when I quit. When I came back, my strength had not gone down much, but computers had shot everyone ahead. My bullet rating peaked at 1601 in 2004. Now it peaks in the low-to-mid 1900s (in a very tough rating pool where many IMs and GMs are rated lower than me), or about 325 points in two years. I lost a good 400 points during the layoff because other players improved. >> I could come back, get a 2300 performance rating, and people would see >> that >> as 500 points below where I need to go. I could be right on schedule to >> get >> there, but it wouldn't matter. For that privilege, I could waste time >> and >> entry fee, instead of staying home and working on my game. > > It's your decision. My decision is, that I simply don't believe that > you gained this strength. I'm crushed. >> > Sure, but there are only few concepts compared to what a GM knows. And >> > a >> > GM can find a move even without calculation, just based on positional >> > knowledge. >> >> Which is how it's possible to "outcalculate" a computer. > > Well, this is the hope. Eventually, they could at least match the computer. >> >Computers are not able to do so. They have to calculate. >> >> So do humans. > > But not to this extent. Computers can do *nothing* without calculation. > But as they can calculate like hell, this results in monsters on the > board. Humans can not calculate so fast and precise as computers, so > the have to rely on other factors. That they can improve this so much, > that they can stay ahead of the computers, is just a hope and there is > not much evidence, that this is the case. I've been able to adapt the computer's concept of positional play (based on calculation) into my own. After a while, I don't need to look more than a few moves ahead to see what the computer would do. The computer just does it every time without fail. >> > Yes, but a GM could make such a move based on general assumptions. He >> > doesn't have to calculate 10 or 15 moves. >> >> How have the "general assumptions" performed so far for players like >> Adams >> against Hydra? > > Not impressive. That's exactly what I am saying: the computers' style > is completely different from a GM. And it doesn't look like the GMs > will keep the edge much longer over those number crunchers. They don't have the edge now. >> >Again: a computer has only very small knowledge about positions. Only if >> >you combine this with his ability to calculate a lot of variations, you >> >get >> >the strength we can see. If you would take the existing evaluations and >> >stop the program thinking beyond move 3, you are back to the 80s with >> >programs just good enough for amateurs. >> >> In the 1980s, Deep Thought was rated 2265. > > Yes, amateur strength. But I meant more those machines which only > calculated 3 moves. Deep Thought calcualted much deeper and was > singular at his time. 2265 is master strength. >> 1956 is "mediocre?" > > Well, I don't juged by numbers of people in the classes, so I am not > talking about average, but by drawing a line between the top and the > beginners. 1900-2000 is somewhere in between. So yes, I regard 1956 as > somewhat mediocre (and my 2150 Elo as well). We're both in the twilight rating area where we can beat most weak players and lose to most strong ones. >> > I don't get it. First: I didn't say anything about the importance, I >> > just >> > said, my rating is not very high. Besides that I don't think, that the >> > rating is actually so important in a debate. >> >> So I have to prove myself, but you can get away with already being weaker >> than me yet still with a better grasp of chess theory than me > > I never claimed to have a better graps of chess theory than you. To the > contrary I would assume, that your knowledge is much deeper. I just win > some games, because I am a fighter, not because I am rekable in > chess theory. You also may play well within your limitations without trying to expand them, as that produces a higher rating, like if someone goes from the King's Gambit to the Ruy, they'll lose some rating points during the learning curve. I like to tear my entire game apart when it levels off at any rating, figuring the structure is unsound, and build a new structure. When I do this, I usually drop 100-200 points for a few weeks, then hit a new peak. >> > My rating doesn't tell you whether I am weaker or stronger as you are. >> >> So you're proud of clouding what is supposed to be an open debate? > > What open debate? Whether you are stonger than me? I simply don't care. > You were about my class when you stopped playing tournaments, now after > so much effort you put into chess, I would assume, that you are much > stronger than me. How much, I don't know. I had to catch up 15 years of computer-fueled theory. I'm also a publisher first and a player second now, so how I do over the board really won't matter to my survival, though winning would be a nice plus. >> My original post merely pointed out that GM play is still weak enough for >> a >> dedicated, talented player to come in and dominate. > > I understood this. But just the logic you based this point on is flawed > in my opinion. Not as flawed as if the humans were so strong they could force draws against the monsters. -- "Google maintains the USENET." -- The Honorable R. Barclay Surrick, Eastern District of PA Judge From Parker v. Google, E.D.Pa. #04-cv-3918
|
| | |
Date: 29 Apr 2006 06:27:23
From: Ralf Callenberg
Subject: Re: There is PLENTY of room at the top in chess
|
29.04.2006 00:05, Ray Gordon: >> And if one is handling long and complicated variations it >> is nearly sure that he will make errors, so it doesn't make sense to go >> any further. [...] > > The Mott Street Chicken doesn't make errors in tic-tac-toe, either, and will > play anyone to a draw or win. The complexity of the two games is not comparable. You see that GMs fail, and this is not because of bad preparation, but because of their limitations. >> No, for stronger players it just takes longer, but even GMs have their >> limits. You can not keep endless long variations in memory and still >> visualize all positions with the same clarity. > > One need not see "all" positions and moves, just the key ones. Oh, this is just a joke like "I only calculate one move, but the right one". You yourself gave examples of how computers broke the limitations of our chess knowledge. And they can do so, because they also examine all those variations a GM would discard because he didn't regard them as "key ones". > >> No. It is simply impossible. The human brain is not capable of >> performing such tasks in this perfection. > > So far, neither is a chess computer. A computer never mixes up variations. He never forgets a position he has examined. For him a position in a variation 10 moves down a side line is as clear as the current position. Humans make mistakes even on those elementary level. GMs later than lay men, but at one point they have to give in. And this is way before computers stop calculating. > > >>>> At least up to now there is no hint into this direction. On the >>>> combinatorical side, the computers' advantage is still increasing. >>> Over the *average* human. >> Also over GMs. > > Over *today's* GMs. First you say *average* human, now you are talking about some future GMs. Discussing with you is quite tedious as you constantly switch you line of argumentation. >> So? He can play it, because he is strong enough. > > IT plays it because IT doesn't understand a) that the pawn isn't really > hanging; and b) that it's antipositional to do so. Can you prove it, if you play against him? > > Karpov broke down physically and mentally. I watched it happen. Had you > been born yet when this match took place? I was following it when it happened. And Karpow was exhausted during the first game. Well, and those break downs are human. They will always happen. So, talking about perfection is just a fairy tale. >> Sure pattern recoginition is important. But with it alone you simply >> can not play. You have to sit down and calculate the variations. >> Pattern recognition is of help, but it does not take all the burdens >> from you. > > Without knowledge of the patterns (like the smothered mate, for example), > calculation becomes much more difficult. Sure. But it's not all. That's what I say. You have to rely on concrete calculation of variations. > >>> That means it's *possible* for a human to exploit this. >> No, your logic is flawed. It only shows, that it is theoretically >> possible to play better than todays computers. > > Right: possible. See above. What "above"? > >> This is absolutely no >> evidence for the claim, that it must be possible for humans to reach >> the same strength without calculation. > > How the human does it isn't relevant. Man, it was you who said it was possible to come over the computers without the calculation power. And I say: your claim has no point. Computers reach their strength by calculation, calculation and only calculation. That humans can surpass them by anything is simply not clear, they are no match when it comes to calculation. And definitely just by the fact, that computers aren't perfect can't you deduce that humans must be able to beat them - with which technique whatsoever. > > Some are. 1. e4 is perfect through move 1. There is still no prove for this. > >> OK. Did you ever switch off the opening theory for Fritz and played >> against him, or one of those new monsters, which you might not know and >> are not prepared for? > > I'm prepared for any opponent at any time. When I lose games, it's not > because of the opening. Well give it a try. Some of those monsters you can download for free. Take one, you don't know and so you didn't explicitly prepare for it. Switch off opening theory, so that he doesn't play the well known lines and play it. >> Of course not. You spent thousands of hours with training and you don't >> have the time and the few buckt to spend a few days for a tournament? > > Why would I bother? I'll play when I'm confident I can win an under-2000 > section, a small open tournament, or grab a GM or IM norm. Until then your claim about the strength you gained has no basis and is just your imagination. That's all what I say. >> Yes, amateur strength. But I meant more those machines which only >> calculated 3 moves. Deep Thought calcualted much deeper and was >> singular at his time. > > 2265 is master strength. Well, it's still amateurs. >>> 1956 is "mediocre?" >> Well, I don't juged by numbers of people in the classes, so I am not >> talking about average, but by drawing a line between the top and the >> beginners. 1900-2000 is somewhere in between. So yes, I regard 1956 as >> somewhat mediocre (and my 2150 Elo as well). > > We're both in the twilight rating area where we can beat most weak players > and lose to most strong ones. And that's what I call mediocre. Greetings, Ralf
|
|
Date: 27 Apr 2006 21:52:18
From: help bot
Subject: Re: There is PLENTY of room at the top in chess
|
Ralf Callenberg wrote: "So you think, from the fact, that computers have shown, that there is a gap between GMs and perfection, follows that humans must be able to enter this gap?" That was Ray Gordon's position, not mine IMO, computers will simply *widen* the gap between themselves and humans. It is a well-known fact that hardware improvements, just as they have in the past, are continuing to leave those of us with just one "processor" and imperfect "memory" in the dust. IMO, Gordon's idea amounts to wishful thinking. What he had right was the part about headroom, that there is a substantial gap between top human GMs and perfect play. Just as there is a gigantic gorge between my play and that of top GMs. :) -- help bot
|
| |
Date: 28 Apr 2006 09:14:38
From: Ralf Callenberg
Subject: Re: There is PLENTY of room at the top in chess
|
28.04.2006 06:52, help bot: > Ralf Callenberg wrote: > > "So you think, from the fact, that computers have shown, that there > is a > gap between GMs and perfection, follows that humans must be able to > enter this gap?" > > > That was Ray Gordon's position, not mine I was asking because you said: > IMO, Ray Gordon has it right. > What he had right was the part about headroom, that there is a > substantial gap between top human GMs and perfect play. Just as there > is a gigantic gorge between my play and that of top GMs. :) Did anybody deny this? The argument goes about the claim, that the gap of the computers over humans shows that human players easily can enter this gap. "if Fritz can do it, so can humans". Greetings, Ralf
|
| |
Date: 28 Apr 2006 01:16:26
From: Ray Gordon
Subject: Re: There is PLENTY of room at the top in chess
|
> "So you think, from the fact, that computers have shown, that there > is a > gap between GMs and perfection, follows that humans must be able to > enter this gap?" > > > That was Ray Gordon's position, not mine IMO, computers will simply > *widen* the gap between themselves and humans. It is a well-known fact > that hardware improvements, just as they have in the past, are > continuing to leave those of us with just one "processor" and imperfect > "memory" in the dust. IMO, Gordon's idea amounts to wishful thinking. My theory is based on the idea that humans learn from mistakes. Search depth is only one aspect of computer play. > What he had right was the part about headroom, that there is a > substantial gap between top human GMs and perfect play. Just as there > is a gigantic gorge between my play and that of top GMs. :) Yep. And only one proven way to close that gap. -- "Google maintains the USENET." -- The Honorable R. Barclay Surrick, Eastern District of PA Judge From Parker v. Google, E.D.Pa. #04-cv-3918
|
| | |
Date: 28 Apr 2006 18:16:21
From: James
Subject: Re: There is PLENTY of room at the top in chess
|
Ray Gordon wrote : >> "So you think, from the fact, that computers have shown, that there >> is a >> gap between GMs and perfection, follows that humans must be able to >> enter this gap?" >> >> >> That was Ray Gordon's position, not mine IMO, computers will simply >> *widen* the gap between themselves and humans. It is a well-known fact >> that hardware improvements, just as they have in the past, are >> continuing to leave those of us with just one "processor" and imperfect >> "memory" in the dust. IMO, Gordon's idea amounts to wishful thinking. > > My theory is based on the idea that humans learn from mistakes. Search > depth is only one aspect of computer play. > Interesting, but quite not the point. Improvements in computer chess have come both from software and hardware. As software is human work, it is likely that improvements will slow down when the field is explored. (But there is still plenty of work to be done in chess). On the opposite, the end of hardware improvements is not near, and running faster makes better programs. Regarding human beings, there are no hardware improvements possible (well, at least not now :-) ) and thus the gap will just become wider. The matter has been settled in checkers with chinook, id for reversi/othello with logistello. Hoping that human beings can always learn enough from their mistakes is purely wishful thinking. An excellent example is Othello. Computers can find the solution of the game something like 25 moves (maybe more now) before the end. No human player can even hope to dream to accomplish that, however hard he would train. Just not enough computation power. And "intuition" is far from being enough to distinguish a won game from a lost game 25 moves before the end. To learn from your mistakes, you have to know what the mistake was, and how to correct it. There, the mistake is just that you are not able to compute fast enough. What have you learned? Nothing useful. You knew it already, and can't do anything about it. James
|
| | | |
Date: 28 Apr 2006 18:07:43
From: Ray Gordon
Subject: Re: There is PLENTY of room at the top in chess
|
> Hoping that human beings can always learn enough from their mistakes is > purely wishful thinking. An excellent example is Othello. > Computers can find the solution of the game something like 25 moves (maybe > more now) before the end. No human player can even hope to dream to > accomplish that, however hard he would train. Just not enough computation > power. And "intuition" is far from being enough to distinguish a won game > from a lost game 25 moves before the end. > To learn from your mistakes, you have to know what the mistake was, and > how to correct it. There, the mistake is just that you are not able to > compute fast enough. What have you learned? Nothing useful. You knew it > already, and can't do anything about it. Sure I can. I can figure out what I did wrong and what I should have done instead, just as I do against humans. -- "Google maintains the USENET." -- The Honorable R. Barclay Surrick, Eastern District of PA Judge From Parker v. Google, E.D.Pa. #04-cv-3918
|
| | | | |
Date: 29 Apr 2006 02:02:53
From: James
Subject: Re: There is PLENTY of room at the top in chess
|
Ray Gordon a �crit : >> Hoping that human beings can always learn enough from their mistakes is >> purely wishful thinking. An excellent example is Othello. >> Computers can find the solution of the game something like 25 moves (maybe >> more now) before the end. No human player can even hope to dream to >> accomplish that, however hard he would train. Just not enough computation >> power. And "intuition" is far from being enough to distinguish a won game >> from a lost game 25 moves before the end. >> To learn from your mistakes, you have to know what the mistake was, and >> how to correct it. There, the mistake is just that you are not able to >> compute fast enough. What have you learned? Nothing useful. You knew it >> already, and can't do anything about it. > > Sure I can. I can figure out what I did wrong and what I should have done > instead, just as I do against humans. > > No you cant't. Let's be more precise. You are at move 35 (25 remainings) in an othello game. There are 15 possible moves, 14 lose, one wins. To know which move wins, you have to compute the game till the end, no other way. You are a human, can not compute, choose the bad move, and lose. However, you don't even know that you have chosen the bad move at this ply. The position may have been lost one, two or more moves before this one. How do you find out? And then, how do you know where the mistake is? Now, let's hope you have a computer to replay the game which shows you the bad move. To be able to learn from your mistake, you must infer a rule in order to avoid this kind of mistake in the future. But even if you memory is perfect, you can only remember that position and its outcome, but this won't tell you anything on the outcome of other very similar positions. The simple flip of one pawn changes the position outcome completely. Thus you have only learned how not to lose a position that certainly won't happen again in your lifetime of Othello player. Information gained: zero. This problem is closely linked to the theory of complexity in computer science. If you have a function such as f(x)= x mod 2 and I ask you to compute that function, you can do it without any problem, for any number. Now, if I ask you to find a subset of the set {23, 45 , 56 , 78, 89} such that the sum of the elements of the subset is 168, it is going to take you a little more time. And the time required grows exponentially with the number of elements, and it has even been proved that there is no way to find a faster algorithm to solve the problem (I apologize to the specialists of complexity theory for these approximations). Computing the endgame of Othello is not conceptually different from the above problem. There are games which have been proved to be PSPACE complete. This means in simple words that there is absolutely no way to find a faster algorithm than pure computation to find the outcome of the game in a general position, and these computational algorithms are themselves "slow" (they require a lot of operations to find the right move). Finding a faster algorithm (which is what you pretend to do by "learning by mistakes", which is in itself an algorithm based on pattern search) is just violating a mathematical theorem (for endgames). Sorry to destroy your hopes. There are two articles on this subject here: http://www.chess-lovers.org/articles/games.pdf http://www.chess-lovers.org/articles/amazons.pdf
|
| | | | | |
Date: 28 Apr 2006 20:38:41
From: Ray Gordon
Subject: Re: There is PLENTY of room at the top in chess
|
>>> Hoping that human beings can always learn enough from their mistakes is >>> purely wishful thinking. An excellent example is Othello. >>> Computers can find the solution of the game something like 25 moves >>> (maybe more now) before the end. No human player can even hope to dream >>> to accomplish that, however hard he would train. Just not enough >>> computation power. And "intuition" is far from being enough to >>> distinguish a won game from a lost game 25 moves before the end. >>> To learn from your mistakes, you have to know what the mistake was, and >>> how to correct it. There, the mistake is just that you are not able to >>> compute fast enough. What have you learned? Nothing useful. You knew it >>> already, and can't do anything about it. >> >> Sure I can. I can figure out what I did wrong and what I should have >> done instead, just as I do against humans. >> >> > No you cant't. The only way to prove that something can't be done is to first assume it can be done and attempt to do it. > Let's be more precise. You are at move 35 (25 remainings) in an othello > game. There are 15 possible moves, 14 lose, one wins. > To know which move wins, you have to compute the game till the end, no > other way. You are a human, can not compute, choose the bad move, and > lose. How did the human get into this position in the first place? Sounds pretty weak. > However, you don't even know that you have chosen the bad move at this > ply. The position may have been lost one, two or more moves before this > one. How do you find out? And then, how do you know where the mistake is? I assume players know how to do this. If not, reserve me a spot at the world chess championship in about eight years. > Now, let's hope you have a computer to replay the game which shows you the > bad move. To be able to learn from your mistake, you must infer a rule in > order to avoid this kind of mistake in the future. This is how chess theory has developed since around 1492. > But even if you memory is perfect, you can only remember that position and > its outcome, but this won't tell you anything on the outcome of other very > similar positions. It doesn't have to. I learned *something* that I wouldn't have learned were it not for the computer. I "solved" part of the game. >The simple flip of one pawn changes the position outcome completely. And there is a way to find this, or we wouldn't know it to begin with. > Thus you have only learned how not to lose a position that certainly won't > happen again in your lifetime of Othello player. Othello has pawns? >Information gained: zero. Not zero, as the game played is information. > This problem is closely linked to the theory of complexity in computer > science. > > If you have a function such as f(x)= x mod 2 > and I ask you to compute that function, you can do it without any problem, > for any number. > Now, if I ask you to find a subset of the set {23, 45 , 56 , 78, 89} such > that the sum of the elements of the subset is 168, it is going to take you > a little more time. And the time required grows exponentially with the > number of elements, and it has even been proved that there is no way to > find a faster algorithm to solve the problem (I apologize to the > specialists of complexity theory for these approximations). Chess pieces, however, are like an army, and there are only 32 of them on a 64-square board, a very small universe. With proper pruning, you can eliminate all but up to six moves per position in 99 percent of the cases, within seconds. > Computing the endgame of Othello is not conceptually different from the > above problem. There are games which have been proved to be PSPACE > complete. This means in simple words that there is absolutely no way to > find a faster algorithm than pure computation to find the outcome of the > game in a general position, and these computational algorithms are > themselves "slow" (they require a lot of operations to find the right > move). A human player who doesn't make mistakes will still draw. The key is using computers to learn how not to make mistakes. If anything, the machines will have more difficulty improving at that level because they will already be calculating incredibly fast. Once you get past looking 10-12 moves ahead, each extra ply requires an incredible amount of computing power. > Finding a faster algorithm (which is what you pretend to do by "learning > by mistakes", which is in itself an algorithm based on pattern search) is > just violating a mathematical theorem (for endgames). > > > Sorry to destroy your hopes. My hope of what? I'm talking theoretically. In real life, I'm improving at the rate of 100-150 ELO points per year. The only long-term battle I have with chess has to do with age, not anything I'll see over the board. In the post-computer era, we can learn from a source that is stronger than the best player in the world. This has never before been the case in chess, so it's logical that what will make players champions in the future will be far different from what made them champions in the past. -- "Google maintains the USENET." -- The Honorable R. Barclay Surrick, Eastern District of PA Judge From Parker v. Google, E.D.Pa. #04-cv-3918
|
| | | | | | |
Date: 29 Apr 2006 13:01:26
From: James
Subject: Re: There is PLENTY of room at the top in chess
|
Ray Gordon a �crit : >>>> Hoping that human beings can always learn enough from their mistakes is >>>> purely wishful thinking. An excellent example is Othello. >>>> Computers can find the solution of the game something like 25 moves >>>> (maybe more now) before the end. No human player can even hope to dream >>>> to accomplish that, however hard he would train. Just not enough >>>> computation power. And "intuition" is far from being enough to >>>> distinguish a won game from a lost game 25 moves before the end. >>>> To learn from your mistakes, you have to know what the mistake was, and >>>> how to correct it. There, the mistake is just that you are not able to >>>> compute fast enough. What have you learned? Nothing useful. You knew it >>>> already, and can't do anything about it. >>> Sure I can. I can figure out what I did wrong and what I should have >>> done instead, just as I do against humans. >>> >>> >> No you cant't. > > The only way to prove that something can't be done is to first assume it can > be done and attempt to do it. Wrong. This is a major mathematical mistake. There are lot of situations in mathematics , (even in game theory) where you can prove a thing without having to try it. The most well know examples are results in cryptography theory where you can prove an algorithm is safe (can not be broken) without having to try to break it (and that's the main interest of the proof). And we are all using the result of this proof when we surf on secure sites on the web. > >> Let's be more precise. You are at move 35 (25 remainings) in an othello >> game. There are 15 possible moves, 14 lose, one wins. >> To know which move wins, you have to compute the game till the end, no >> other way. You are a human, can not compute, choose the bad move, and >> lose. > > How did the human get into this position in the first place? Sounds pretty > weak. The human might have got in this position by perfect play. A position with one (and only one) winning move is still a won position. It might very well be that things were perfect till there. > >> However, you don't even know that you have chosen the bad move at this >> ply. The position may have been lost one, two or more moves before this >> one. How do you find out? And then, how do you know where the mistake is? > > I assume players know how to do this. If not, reserve me a spot at the > world chess championship in about eight years. Still wrong, they can't and endgames are just a blatant example. If you are unfamiliar with Othello, think about six pieces endgame tables in chess. The computer can play perfectly, while you will never be able to, however hard you try and this is a fact; endgame tables can not be much "compressed" (regarding chaitin's complexity), thus you won't be able to learn them except by remembering them all: no rule infering is possible. Can you remember terabytes of data? The computer can. > >> Now, let's hope you have a computer to replay the game which shows you the >> bad move. To be able to learn from your mistake, you must infer a rule in >> order to avoid this kind of mistake in the future. > > This is how chess theory has developed since around 1492. > This kind of learning is limited by results of game theory (and theory here is used in the mathematical sense, nothing like chess theory...) which applies equally to computers, human beings, tians and you. Have a look at Rybka's homepage if you want to read what its author (which is a strong chess player also) thinks about the truth of chess theory. Chess theory is a collection of rules of thumb made by human beings for human beings in order to be able to find a good move. Quite strangely, your belief is the dual of an old belief of the 70's in the AI community that using human knowledge would increase the strength of programs (the old expert systems vs brute force debate). That debate is over in our community for years now, and the thing we have all learned is that human knowledge and computing don't mix up well (read Schaeffer or Berliner on that). > >> But even if you memory is perfect, you can only remember that position and >> its outcome, but this won't tell you anything on the outcome of other very >> similar positions. > > It doesn't have to. I learned *something* that I wouldn't have learned were > it not for the computer. I "solved" part of the game. > >> The simple flip of one pawn changes the position outcome completely. > > And there is a way to find this, or we wouldn't know it to begin with. Yes, you can find this out with a computer, no other way, because you must compute to find out. That's how we know. >> Thus you have only learned how not to lose a position that certainly won't >> happen again in your lifetime of Othello player. >> Information gained: zero. > > Not zero, as the game played is information. > Well information gained regarding endgame positions is 1/(2^60) and this is probably a large upper bound. Close to zero, don't you think... > >> This problem is closely linked to the theory of complexity in computer >> science. >> >> If you have a function such as f(x)= x mod 2 >> and I ask you to compute that function, you can do it without any problem, >> for any number. >> Now, if I ask you to find a subset of the set {23, 45 , 56 , 78, 89} such >> that the sum of the elements of the subset is 168, it is going to take you >> a little more time. And the time required grows exponentially with the >> number of elements, and it has even been proved that there is no way to >> find a faster algorithm to solve the problem (I apologize to the >> specialists of complexity theory for these approximations). > > Chess pieces, however, are like an army, and there are only 32 of them on a > 64-square board, a very small universe. Wrong again. 32 pieces on a 64 square board is a gigantic universe. Even much more simple games than chess are PSPACE complete. Dont't you remember the story of the chinese wise man asking as a reward that his lord gives him 1 bean on the first square of the chess board, 2 on the second, 4 on the third and so on? You can make very big numbers with very small ones... > With proper pruning, you can eliminate all but up to six moves per position > in 99 percent of the cases, within seconds. Not much of a performance. In chess, any alpha-beta algorithm properly used (and without extra pruning technique) won't search more than 5 moves on the average at each ply (well technically it will search on the whole a number of leaves which is something like the square root of the branching factor raised to the power of the number of plies searched). > >> Computing the endgame of Othello is not conceptually different from the >> above problem. There are games which have been proved to be PSPACE >> complete. This means in simple words that there is absolutely no way to >> find a faster algorithm than pure computation to find the outcome of the >> game in a general position, and these computational algorithms are >> themselves "slow" (they require a lot of operations to find the right >> move). > > A human player who doesn't make mistakes will still draw. > > The key is using computers to learn how not to make mistakes. > > If anything, the machines will have more difficulty improving at that level > because they will already be calculating incredibly fast. Once you get past > looking 10-12 moves ahead, each extra ply requires an incredible amount of > computing power. Wrong. It requires exactly the same increase in power to go from 4 to 5 than to 12 to 13, no more, no less (approximatively the square root of the branching factor, for any game, without extra pruning). > >> Finding a faster algorithm (which is what you pretend to do by "learning >> by mistakes", which is in itself an algorithm based on pattern search) is >> just violating a mathematical theorem (for endgames). >> >> >> Sorry to destroy your hopes. > > My hope of what? I'm talking theoretically. > In real life, I'm improving at the rate of 100-150 ELO points per year. The > only long-term battle I have with chess has to do with age, not anything > I'll see over the board. I long to see that. Keep us informed. However there is a rule of thumb regarding the learning curve of anything (and I say a rule not a proof here, it's much different) which states that learning is only linear at the start of the curve and becomes asymptotic afterwards. Can you tell us which is you current rating? We will check out next year, and the following ones. An interesting experiment indeed. > In the post-computer era, we can learn from a source that is stronger than > the best player in the world. This has never before been the case in chess, > so it's logical that what will make players champions in the future will be > far different from what made them champions in the past. > The fact that electronic calculators compute extremely fast hasn't made human beings faster at computing (quite the opposite in fact, as most of them have stopped to compute at all...). There is nothing to learn from the way your calculator multiplies two large numbers in order to increase your own computing skill.
|
| | | | | | | |
Date: 30 Apr 2006 11:07:54
From: Ray Gordon
Subject: Re: There is PLENTY of room at the top in chess
|
>>> No you cant't. >> >> The only way to prove that something can't be done is to first assume it >> can be done and attempt to do it. > > Wrong. This is a major mathematical mistake. There are lot of situations > in mathematics , (even in game theory) where you can prove a thing without > having to try it. The most well know examples are results in cryptography > theory where you can prove an algorithm is safe (can not be broken) > without having to try to break it (and that's the main interest of the > proof). And we are all using the result of this proof when we surf on > secure sites on the web. Yet that doesn't apply here, and hackers make mockeries of technology all the time. The Scientific Method says tests are necessary as a rule, and that they be reproducible. >>> To know which move wins, you have to compute the game till the end, no >>> other way. You are a human, can not compute, choose the bad move, and >>> lose. >> >> How did the human get into this position in the first place? Sounds >> pretty weak. > > The human might have got in this position by perfect play. Where 14 of 15 moves lose? >A position with one (and only one) winning move is still a won position. It >might very well be that things were perfect till there. Perhaps. Still, something separates that winning move, which can therefore be found. >>> However, you don't even know that you have chosen the bad move at this >>> ply. The position may have been lost one, two or more moves before this >>> one. How do you find out? And then, how do you know where the mistake >>> is? >> >> I assume players know how to do this. If not, reserve me a spot at the >> world chess championship in about eight years. > > Still wrong, they can't and endgames are just a blatant example. If you > are unfamiliar with Othello, think about six pieces endgame tables in > chess. The computer can play perfectly, while you will never be able to, I'll never need to. > however hard you try and this is a fact; endgame tables can not be much > "compressed" (regarding chaitin's complexity), thus you won't be able to > learn them except by remembering them all: no rule infering is possible. > Can you remember terabytes of data? The computer can. Computers can't find the best moves in the opening or middlegame so far. The more I use them, the more I see their weaknesses firsthand. >>> Now, let's hope you have a computer to replay the game which shows you >>> the bad move. To be able to learn from your mistake, you must infer a >>> rule in order to avoid this kind of mistake in the future. >> >> This is how chess theory has developed since around 1492. >> > > This kind of learning is limited by results of game theory (and theory > here is used in the mathematical sense, nothing like chess theory...) > which applies equally to computers, human beings, tians and you. > Have a look at Rybka's homepage if you want to read what its author (which > is a strong chess player also) thinks about the truth of chess theory. > Chess theory is a collection of rules of thumb made by human beings for > human beings in order to be able to find a good move. Pretty much. > Quite strangely, your belief is the dual of an old belief of the 70's in > the AI community that using human knowledge would increase the strength of > programs (the old expert systems vs brute force debate). That debate is > over in our community for years now, and the thing we have all learned is > that human knowledge and computing don't mix up well (read Schaeffer or > Berliner on that). Or "garbage in, garbage out" has been the result. >>> But even if you memory is perfect, you can only remember that position >>> and its outcome, but this won't tell you anything on the outcome of >>> other very similar positions. >> >> It doesn't have to. I learned *something* that I wouldn't have learned >> were it not for the computer. I "solved" part of the game. >> >>> The simple flip of one pawn changes the position outcome completely. >> >> And there is a way to find this, or we wouldn't know it to begin with. > > Yes, you can find this out with a computer, no other way, because you must > compute to find out. That's how we know. That only applies to positions I haven't solved. >>> Thus you have only learned how not to lose a position that certainly >>> won't happen again in your lifetime of Othello player. Information >>> gained: zero. >> >> Not zero, as the game played is information. >> > > Well information gained regarding endgame positions is 1/(2^60) and this > is probably a large upper bound. Close to zero, don't you think... See my 1/x rule. The endgame to me is just a broken play from the opening, like in the NFL when the QB is out of the pocket. >> Chess pieces, however, are like an army, and there are only 32 of them on >> a 64-square board, a very small universe. > > Wrong again. 32 pieces on a 64 square board is a gigantic universe. Not as gigantic as Go. >Even much more simple games than chess are PSPACE complete. > Dont't you remember the story of the chinese wise man asking as a reward > that his lord gives him 1 bean on the first square of the chess board, 2 > on the second, 4 on the third and so on? You can make very big numbers > with very small ones... You can also use rules of thumb to prune. >> With proper pruning, you can eliminate all but up to six moves per >> position in 99 percent of the cases, within seconds. > > Not much of a performance. In chess, any alpha-beta algorithm properly > used (and without extra pruning technique) won't search more than 5 moves > on the average at each ply (well technically it will search on the whole a > number of leaves which is something like the square root of the branching > factor raised to the power of the number of plies searched). Chess computers are not very good at pruning, from what I've seen. They leave out many good moves and get "surprised" a great deal. >> A human player who doesn't make mistakes will still draw. >> >> The key is using computers to learn how not to make mistakes. >> >> If anything, the machines will have more difficulty improving at that >> level because they will already be calculating incredibly fast. Once you >> get past looking 10-12 moves ahead, each extra ply requires an incredible >> amount of computing power. > > Wrong. It requires exactly the same increase in power to go from 4 to 5 > than to 12 to 13, no more, no less (approximatively the square root of the > branching factor, for any game, without extra pruning). I disagree, given that each ply builds on all previous plies. Otherwise, looking 100 moves ahead would be the same as looking one move ahead. >>> Finding a faster algorithm (which is what you pretend to do by "learning >>> by mistakes", which is in itself an algorithm based on pattern search) >>> is just violating a mathematical theorem (for endgames). >>> >>> >>> Sorry to destroy your hopes. >> >> My hope of what? I'm talking theoretically. >> In real life, I'm improving at the rate of 100-150 ELO points per year. >> The only long-term battle I have with chess has to do with age, not >> anything I'll see over the board. > > I long to see that. Keep us informed. However there is a rule of thumb > regarding the learning curve of anything (and I say a rule not a proof > here, it's much different) which states that learning is only linear at > the start of the curve and becomes asymptotic afterwards. Can you tell us > which is you current rating? We will check out next year, and the > following ones. An interesting experiment indeed. My former USCF rating was 1900 when I quit, 2000 at its peak. I plan to play in tournaments when I'm capable of grabbing a GM (or IM) norm, which is really the only reason I'd ever need to do so. Otherwise it's a colossal waste of time and money. I just hit a new peak today in fact, 20 points above my previous peak, which keeps me on schedule for at least the next two months. If you saw how systematic my improvement was, and the way I play the opening (no offense but I'm not going to give away all my secrets just yet), you wouldn't be so quick to declare the man v. machine battle over. Computers already have to wait a very long time to beat me, and that is increasing steadily. I couldn't begin to tell you what these things have taught me about chess. >> In the post-computer era, we can learn from a source that is stronger >> than the best player in the world. This has never before been the case >> in chess, so it's logical that what will make players champions in the >> future will be far different from what made them champions in the past. >> > > The fact that electronic calculators compute extremely fast hasn't made > human beings faster at computing (quite the opposite in fact, as most of > them have stopped to compute at all...). Math is not competitive in that sense. >There is nothing to learn from the way your calculator multiplies two large >numbers in order to increase your own computing skill. Actually there is a lot to learn from equations that couldn't be done by a mere human. For example, thanks to computers I was able to do my NCAA power ratings (based on the Elo system) with 10-20 minutes of data input even on a Saturday (100+ games), and of course the database can calculate all types of ratings that I could never have done by hand. It used to take three hours every Sunday to make those ratings current. -- "Google maintains the USENET." -- The Honorable R. Barclay Surrick, Eastern District of PA Judge From Parker v. Google, E.D.Pa. #04-cv-3918
|
| | | | | | | | |
Date: 01 May 2006 03:50:43
From: James
Subject: Re: There is PLENTY of room at the top in chess
|
Ray Gordon wrote : >> Wrong. This is a major mathematical mistake. There are lot of situations >> in mathematics , (even in game theory) where you can prove a thing without >> having to try it. The most well know examples are results in cryptography >> theory where you can prove an algorithm is safe (can not be broken) >> without having to try to break it (and that's the main interest of the >> proof). And we are all using the result of this proof when we surf on >> secure sites on the web. > > Yet that doesn't apply here, and hackers make mockeries of technology all > the time. > Wrong. That's just legend. Hackers only make mockeries of technology in hollywood movies. Read Mitnick's book "The Art of Deception" (among others)to understand how hackers really operate. They are just making mockeries of human beings (what Mitnick calls social engineering). > The Scientific Method says tests are necessary as a rule, and that they be > reproducible. Wrong in this case. The "scientific method" says many things. We could discuss for hours the empiricist / observationalist / inductivist philosophy, the falsifiability theory of Karl Popper, and Kuhn's paradigmatic theory. Unfortunately, these philosophical discussions have nothing to do with mathematics (and game theory is a part of mathematics), they only apply to induction based experimental theory. You don't have to test a mathematical theorem as soon as it has been proven regarding a set of axioms and inference rules. I am sure you can do better by digging out the formalist/constructivist controversy or one of Goedel's theorem: there are plenty of interesting things in mathematical philosophy. I would really like to see you try... However, your definition is interesting, because it applies perfectly to the sentence "Ray Gordon's rating increases by 100-150 points each year". This sentence needs some tests to be validated, a test like a good old chess tournament each year (at least) for example. Till then, Popper would have told you that it's just on par with astrology, sorcery or parapsychology. >> Quite strangely, your belief is the dual of an old belief of the 70's in >> the AI community that using human knowledge would increase the strength of >> programs (the old expert systems vs brute force debate). That debate is >> over in our community for years now, and the thing we have all learned is >> that human knowledge and computing don't mix up well (read Schaeffer or >> Berliner on that). > > Or "garbage in, garbage out" has been the result. > ??? Do you call Logistello, Chinook or current chess programs "garbage"??? > > Chess computers are not very good at pruning, from what I've seen. They > leave out many good moves and get "surprised" a great deal. > Wrong. You are confusing two meanings of the word pruning. Pruning done by an alpha-beta is always mathematically correct, while some specific heuristics (such as the "null-move" heuristic pruning) can miss some combinations in order to increase search speed. I was speaking about alpha-beta pruning only. Read carefully. > >> Wrong. It requires exactly the same increase in power to go from 4 to 5 >> than to 12 to 13, no more, no less (approximatively the square root of the >> branching factor, for any game, without extra pruning). > > I disagree, given that each ply builds on all previous plies. Otherwise, > looking 100 moves ahead would be the same as looking one move ahead. Wrong. As stated above, the time required to search a position is proportional to the number of positions searched which (again) grows like the square root of the branching factor raised to the power of the number of plies. Thus, in chess, to search 6 plies requires approximately 5 times (without heuristic pruning) the time required to search 5 plies. Id to go from 12 to 13. There is absolutely no difference between 5- >6 and 12->13. So, if Moore's law remains valid, computers will gain automatically one ply each 2-4 years. Basic mathematics. > > My former USCF rating was 1900 when I quit, 2000 at its peak. I plan to > play in tournaments when I'm capable of grabbing a GM (or IM) norm, which is > really the only reason I'd ever need to do so. Otherwise it's a colossal > waste of time and money. I just hit a new peak today in fact, 20 points > above my previous peak, which keeps me on schedule for at least the next two > months. > > If you saw how systematic my improvement was, and the way I play the opening > (no offense but I'm not going to give away all my secrets just yet), you > wouldn't be so quick to declare the man v. machine battle over. Computers > already have to wait a very long time to beat me, and that is increasing > steadily. I couldn't begin to tell you what these things have taught me > about chess. > > When do you plan to have a few real tests to turn your theory into something "scientific"?
|
|
Date: 27 Apr 2006 09:45:59
From: Ralf Callenberg
Subject: Re: There is PLENTY of room at the top in chess
|
Ray Gordon schrieb: > >> My point was that the gap is large enough for a human to easily enter it > >> without even approaching what the computers can do. > > > > And my point is, that this is taken out of thin air. There is also a > > huge gap between Michael Johnson and a cheetah - nevertheless humans > > are simply not able to enter this gap much deeper. > > Humans can learn chess from computers, since the computers have to divulge > the moves they play, if not the reasons behind the move. Watching stronger players always helps. So, now GMs have somebody to watch who is stronger. That's true. GMs might increase their strength due to computer - that's true. But your logic "Fritz can do it so humans can do it" simply is still not a valid conclusion. > Knowing how a cheetah runs so fast will not help a human to run faster, but > learning that your pet opening has a hidden move that gives you a +3.43 > evaluation a few moves down the road definitely will. Yes, opening preparation is supported by computers. But does this increase the playing strength over the board? Here and there yes, but in general - the way computers play is completely different from humans. Humans will never be able to calculate variations with the perfection of computers. > Your example is not relevant to this issue. It is. You are mixing up two different lines of argumentation. First you said: computers showed us that there is a gap, and therefore man can enter this gap. This is still not proven by you and against this I used my examples. The argument, that computer can help to increase the strength of players, came later. This second argument I would not completely dismiss, although I think you are exaggerating the influence of computers. > If the gap were "not very deep" then computers wouldn't be crushing our GMs > because our GMs would already be close to perfect. Hello? Who said, the gap was not very deep? I said, it is not clear, that humans can enter this gap very deep. > > Yes, this happened again and again,. Lasker and Capablanca are examples > > of the past. There will always be one or two dominating their era. And > > what does this tell us? > > That it's possible for someone to dominate his peers in chess, and that the > dreaded "draw death" isn't anywhere close on the horizon. This became obvious some 30 years ago when Capablanca's prophecy didn't come true. So what? I still don't see your point. > > Probably the strength has increased because knowledge has increased, > > In large part because of computers! In some parts. The amount of influence is just speculation. Chess also made advances long before chess computers became reasonable opponents. > > the use of improved techniques. But by which gin the strength has > > increased is a matter of speculation. > > Not really, since we can run the games of every generation through the > silicon to see who knew what. Has this been done? > >And the advance of machines > > doesn't tell us anything about the possibities of the human mind. > > Of course it does! It shows us the ultimate perfection that can be achieved > when humans reduce chess to a set of general principles that allow them to > never lose. The principles computers use are simply not transferable to the human domain. We will never be able to evaluate millions of positions per second. And this ability is the corner stone of the strenght of Fritz & Co. These monsters have shown us, that concrete calculation beats knowledge. How can humans take advantage in a general way from this? > >They > > are simply not related. So your statement "If Fritz can do it, humans > > can do it" is just a claim without the slightest trace of proof for it. > > Hardly: Fritz can easily teach us what is correct to do in a given position. > We can learn openings, positional themes, tactics, etc. from it. The moves > it plays are played for reasons, very sound reasons, and it's up to us to > uncover them. For opening preparations this might be useful. But if the computers have shown us something it is simply: "Stop thinking and calculate!" If you are in a new situation on the board, what computers would do here is of no use for you, as you simply can not calculate the way computers do. > At least I *thought* this was what people were > doing, because if they aren't, you better reserve a seat for me at the world > championship sometime in the next decade since my training has me on a much > faster track. Sure... > > Computers and humans play chess completely different. > As a Eurofighter > > flies completely different than a falcon. > > How computers arrive at their moves is secondary to the fact that their > moves are the ones that the top players need to be studying. But what if the players can not deduce some general principle? If it really comes down to "Calculate, calculate"? So, the way computers get their moves is absolutely not secondary. If you can not transfer their ability to the way humans can think, their way of playing is only of little use. Computers might help us in several insights, but how far this will go is only speculation. > My game has already changed immensely because of what computers taught me, > and I spent 75-80 hours a week training back in the late 1980s, when they > were not reliable training tools (strongest computer was about 2250 and only > reliable in heavily tactical situations at slow time controls). Maybe it was just, because you are a weak player and spending so much time playing chess would have changed your playing anyway - with or without computers. Who knows? You are extraploating from a single case - yours. This is not valid. Greetings, Ralf
|
| |
Date: 27 Apr 2006 14:44:42
From: Ray Gordon
Subject: Re: There is PLENTY of room at the top in chess
|
>> >> My point was that the gap is large enough for a human to easily enter >> >> it >> >> without even approaching what the computers can do. >> > >> > And my point is, that this is taken out of thin air. There is also a >> > huge gap between Michael Johnson and a cheetah - nevertheless humans >> > are simply not able to enter this gap much deeper. >> >> Humans can learn chess from computers, since the computers have to >> divulge >> the moves they play, if not the reasons behind the move. > > Watching stronger players always helps. So, now GMs have somebody to > watch who is stronger. That's true. GMs might increase their strength > due to computer - that's true. But your logic "Fritz can do it so > humans can do it" simply is still not a valid conclusion. Yet the only way to refute my claim is to actually TRY to play as well as a computer. The computers have changed the nature of chess, AND the skill set required to become a world champion. >> Knowing how a cheetah runs so fast will not help a human to run faster, >> but >> learning that your pet opening has a hidden move that gives you a +3.43 >> evaluation a few moves down the road definitely will. > > Yes, opening preparation is supported by computers. But does this > increase the playing strength over the board? Here and there yes, So being booked up perfectly to move 30 is "here and there?" You must not know many video game wizards, because that type of task is child's play to them. >but > in general - the way computers play is completely different from > humans. Is it? >Humans will never be able to calculate variations with the > perfection of computers. That sound you hear is Mikhail Tal laughing from his grave. >> Your example is not relevant to this issue. > > It is. You are mixing up two different lines of argumentation. First > you said: computers showed us that there is a gap, and therefore man > can enter this gap. This is still not proven by you and against this I > used my examples. The argument, that computer can help to increase the > strength of players, came later. You had said that a human can't learn how to run faster by watching a cheetah, and I pointed out that humans can learn how to play chess better by watching Fritz playing with itself. Very relevant. Actually, humans can learn how to run faster by watching animals, and in fact we study their strides for just that purpose. Today's racehorses are often tested the way olympic athletes are to see how favorable their body type is to running. >This second argument I would not > completely dismiss, although I think you are exaggerating the influence > of computers. I trained like a demon without computers, and I'm training again like a demon with them. That I am even bothering to study this game at my age should serve as notice that I see a path to becoming at least a GM and probably well beyond that. If I don't get there, whoever continues my work after I write my books, will. For me, it's just a matter of methodically eliminating my technical weaknesses after move 20. I doubt there is a player in the world who can consistently get an advantage against me out of the opening, and if there is, within two years, there won't be. So far the list of players who have not been able to best me in the opening includes Kamsky (twice) and Nakamura (twice), as well as many other GMs and IMs you've never heard of. This with my playing strength at or below 2000. Of course, there are probably some 1850s who would give me more hassles if they study like I do. The question is, is it easier for those players to get an opening repertoire or for me to get a middlegame and endgame? I've never studied those two parts of the game for very long (maybe six months for each so far, but that will be changing). When I began training, I believed that if I could become 2700 strength in the opening, that I could then turn my attention to the later parts of the game and become that strength in all three phases. I call it "isolation and integration." The reason the opening comes first is so that I don't have to change my repertoire as I get better, and so I have better middlegames and endgames to study, since I know for sure what positions I'll be reaching. In fact, right now I found a line that is all but a forced win for White (the books say it's "unclear" or "even"), with Black's only chance for a draw is a line that leads to an opposite-bishop ending where White is a pawn up and probably winning, but I hate that type of ending. Without my repertoire, I would not know that I need to study this type of ending, and even if I did study opposite-bishop endings (which I'm actually pretty good at), that wouldn't cover the specific puzzle with which I'm confronted. Adams didn't even make it to the endgame against Hydra much, now did he? What computers do is simple: they play main-line openings until about move 10-12, then they look for a line that isn't well studied, play it to take the human out of book, and crush it tactically. You think a human can't adopt a similar approach? >> If the gap were "not very deep" then computers wouldn't be crushing our >> GMs >> because our GMs would already be close to perfect. > > Hello? Who said, the gap was not very deep? I said, it is not clear, > that humans can enter this gap very deep. Most top GMs would probably tell you that there will come a day when a human can play perfect chess, or very close to perfect. Before computers, it was believed that this was the case, and that we were already there (the "draw death" for example). Using the "bear" example, if the computer is the "bear," I don't have to outrun it to survive, I only have to outrun YOU. >> > Yes, this happened again and again,. Lasker and Capablanca are examples >> > of the past. There will always be one or two dominating their era. And >> > what does this tell us? >> >> That it's possible for someone to dominate his peers in chess, and that >> the >> dreaded "draw death" isn't anywhere close on the horizon. > > This became obvious some 30 years ago when Capablanca's prophecy didn't > come true. So what? I still don't see your point. Kasparov-Karpov I was called off because of the ridiculous number of draws. I still hear people whining about it now. >> > Probably the strength has increased because knowledge has increased, >> >> In large part because of computers! > > In some parts. The amount of influence is just speculation. Chess also > made advances long before chess computers became reasonable opponents. Kasparov said pretty much what I did, and admitted that he used them all the time in his training. Leko has also spoken of it. Kasparov actually referred to the pre-computer era in chess as the "dark ages." >> > the use of improved techniques. But by which gin the strength has >> > increased is a matter of speculation. >> >> Not really, since we can run the games of every generation through the >> silicon to see who knew what. > > Has this been done? Many, many times, usually with the person doing it breaking out in laughter at all the punctuation after the various moves in the annotation. Let's just say that Fischer's 60 memorable games aren't so memorable anymore. >> >And the advance of machines >> > doesn't tell us anything about the possibities of the human mind. >> >> Of course it does! It shows us the ultimate perfection that can be >> achieved >> when humans reduce chess to a set of general principles that allow them >> to >> never lose. > > The principles computers use are simply not transferable to the human > domain. Of course they are. Every move has a reason behind it, and that reason can be reverse-engineered. I learn something new every time I watch a computer play. >We will never be able to evaluate millions of positions per > second. We will never have to. We have the ability to create shortcuts that the computers do not. What the computer sees tactically, we can figure out positionally. Want an example? Time is the most overrated factor in the opening. Want to see a practical example of this? Try this: 1. e4 e6 2. f4 Qh4+!!! 3. g3 Qd8!! Is Black busted? Not even close. In fact, he's just played the BEST move each time. How can that be? Black just LOST TWO TEMPO IN THE OPENING! Check your engines, however, and see if you can force an advantage, let alone a win. Before computers, we just ASSUMED that it had to be bad to do something like that, and even if we couldn't find the refutation we just ASSUMED that it existed, somewhere, for some GM to kick our ass with at the worst possible time. In this particular case, putting White's pawn at g3 is well worth the loss of two moves. Now WHY is this the case? There is a general principle here that explains this, and which also explains that time is the most overrated factor in the opening. >And this ability is the corner stone of the strenght of Fritz & > Co. These monsters have shown us, that concrete calculation beats > knowledge. Or they have shown us that inferior play in the opening is punished tactically. What if the computer never gets anything to sink its tactical teeth into, not because the human locked up the position, but because he played PERFECTLY? I've already scored wins over those 3000-rated supermonsters on ICC, including one against a machine that had over 7,000 wins and 36 losses. I smashed it on the White side of a Sicilian Dragon at one-minute time controls. Granted, they win most of the time, but you have to start somewhere, and many players, even GMs, can't say they've beaten the machine I beat, even once. If there is "room at the top" for computers, there's a goddamn canyon for the humans. >How can humans take advantage in a general way from this? Take advantage of the computer's tactical knowledge? The same way one always improves at chess: PATTERN RECOGNITION. Just like humans, computers have styles and "personalities" all their own. The only difference is that computers don't annotate their games or explain their brilliance to the world; we players have to decipher their move-selection process on their own. >> >They >> > are simply not related. So your statement "If Fritz can do it, humans >> > can do it" is just a claim without the slightest trace of proof for it. >> >> Hardly: Fritz can easily teach us what is correct to do in a given >> position. >> We can learn openings, positional themes, tactics, etc. from it. The >> moves >> it plays are played for reasons, very sound reasons, and it's up to us to >> uncover them. > > For opening preparations this might be useful. But if the computers > have shown us something it is simply: "Stop thinking and calculate!" What they have taught me is to keep the advantage against them from the very first move and never let go of it. My "1/x rule." (that's where from any given position, the next move has an "importance value" of 1, with every subsequent move being worth 1/x). Fact is, once a computer has an edge against you, you're as dead as in a video game where your man is about to be killed off. The trick is to never let the machine get you on the mat. Of course, people who get "creative" in the opening with crap like the Latvian aren't going to see this so easily, and will blame the computer's "tactical superiority" on their losses. Computers will be ushering out the era where people play garbage lines....or will they be ushered IN as a form of complex trap for the unprepared? >If > you are in a new situation on the board, what computers would do here > is of no use for you, as you simply can not calculate the way computers > do. But I can ask: WWFD? (What would FRITZ do?). Pattern recognition can lead one to figure out what a computer would "think" of a position. Tal, probably the greatest tactician of all time, understood this instinctively, and often would sacrifice a knight in the Sicilian just on general principle. Was it sound? Not always, but computer tactics aren't always sound either. Ever hear of the "horizon effect?" It works for tactics as well as for positional play. >> At least I *thought* this was what people were >> doing, because if they aren't, you better reserve a seat for me at the >> world >> championship sometime in the next decade since my training has me on a >> much >> faster track. > > Sure... If no one else is training the way I am now training, I will have little competition in ten years if my body holds up, but that's a big IF. More likely, however, there are young players who are training this way and will realize this vision. Too many st people play this game for me to be the only one. That said, it's actually healthy to ignore the negatives in pursuit of a goal like that, as one business success said, since if you focus on the obstacles, you'll never get anywhere. I actually got back into chess for the purpose of writing a book on how to play, so my style will be in print, hopefully by the summer. Much as I wanted to have the element of surprise in my favor (none of my games are in any database), it is more important to secure an income as a chess author that can further fund my training. On the upside, I'll have readers to help do a lot of my research for me. >> > Computers and humans play chess completely different. >> As a Eurofighter >> > flies completely different than a falcon. >> >> How computers arrive at their moves is secondary to the fact that their >> moves are the ones that the top players need to be studying. > > But what if the players can not deduce some general principle? The phrase "don't quit your day job" comes to mind. >If it > really comes down to "Calculate, calculate"? So, the way computers get > their moves is absolutely not secondary. Computers actually use positional factors to evaluate the result of their "calculation." You think they find moves like Kh1 after castling through brute tactical force? >If you can not transfer their > ability to the way humans can think, their way of playing is only of > little use. See the "don't quit your day job" rek regarding that. Chess will be taken over by players who, through pattern recognition, can decipher this stuff. Further, even those who can't will still have their openings "proofread" by computers, and this has already shown up in the play of many players, who don't get busted very easily in the opening. >Computers might help us in several insights, but how far > this will go is only speculation. Only one way to find out... >> My game has already changed immensely because of what computers taught >> me, >> and I spent 75-80 hours a week training back in the late 1980s, when they >> were not reliable training tools (strongest computer was about 2250 and >> only >> reliable in heavily tactical situations at slow time controls). > > Maybe it was just, because you are a weak player and spending so much > time playing chess would have changed your playing anyway - with or > without computers. Who knows? You are extraploating from a single case > - yours. This is not valid. Weak player? I trained for four years and didn't chase rating points. My average rate of improvement -- then and now -- is a very steady 100-150 elo points per year. I'm still a work in progress, but everyone has to start somewhere. One reason people don't go far over 2000 very often is that they don't build the right foundation for improvement and they lack the patience to endure the brutal losses that are inevitable when one stretches their chess knowledge. Do musicians play perfect symphonys their first attempt? I should further clarify that my improvement in the 1980s was achieved without computers, not with. My recent improvement is almost exclusively the product of the engines. I gather you are a very high-rated player, but really anything under 2700 is just "basic," relatively speaking. Getting from 2700-2800 probably requires as much or more work than getting from 1500-2700. I bet if Kasparov had started late in life rather than early, and trained at full intensity, he would have made it to 2600 or 2700 pretty fast. Whatever the ultimate truth here, no one ever became a chess champion by imitating his peers, and I'm obviously not doing that. -- "Google maintains the USENET." -- The Honorable R. Barclay Surrick, Eastern District of PA Judge From Parker v. Google, E.D.Pa. #04-cv-3918
|
| | |
Date: 28 Apr 2006 00:44:44
From: Ralf Callenberg
Subject: Re: There is PLENTY of room at the top in chess
|
27.04.2006 20:44, Ray Gordon: > Yet the only way to refute my claim I don't have to. YOU made the claim, YOU have to prove it. As long as you don't, I might say, it is just a claim, but nothing is proven. > So being booked up perfectly to move 30 is "here and there?" People are still not perfect to move 30. And a guy like Aronyan shows, that you don't have to prepare your openings like hell to go to 2700. > >> Humans will never be able to calculate variations with the >> perfection of computers. > > That sound you hear is Mikhail Tal laughing from his grave. Indeed, he was successful without perfect calculations. With his typical style of play, Tal wouldn't have too big chances against Fritz. > What computers do is simple: they play main-line openings until about move > 10-12, then they look for a line that isn't well studied, play it to take > the human out of book, and crush it tactically. > > You think a human can't adopt a similar approach? No. Simply because they can not play tactically on this level. > Most top GMs would probably tell you that there will come a day when a human > can play perfect chess, or very close to perfect. Which GM said this for example? > Before computers, it was > believed that this was the case, and that we were already there (the "draw > death" for example). Until when did most people believe this? Capablanca said so in the 30s. 40 years later this was still not the case. > Many, many times, usually with the person doing it breaking out in laughter > at all the punctuation after the various moves in the annotation. Let's > just say that Fischer's 60 memorable games aren't so memorable anymore. Whoever says something like that doesn't know too much about chess. > Of course they are. Every move has a reason behind it, and that reason can > be reverse-engineered. I learn something new every time I watch a computer > play. Yes, but if a concrete situation deserves concrete calculations reverse engineering will only help you in the post mortem analysis. >> We will never be able to evaluate millions of positions per >> second. > > We will never have to. We have the ability to create shortcuts that the > computers do not. What the computer sees tactically, we can figure out > positionally. It is an unproven claim that this is possible in a way much beyond what is possible for humans today. > In this particular case, putting White's pawn at g3 is well worth the loss > of two moves. Now WHY is this the case? There is a general principle here > that explains this, and which also explains that time is the most overrated > factor in the opening. I think there are older examples where one side deliberately gives away a tempo in order to force the other side to make unfavorable pawn moves or to put pieces on unfortunate positions. So this is not such a breathtaking knowledge unheard of before the use of computers. >> And this ability is the corner stone of the strenght of Fritz & >> Co. These monsters have shown us, that concrete calculation beats >> knowledge. > > Or they have shown us that inferior play in the opening is punished > tactically. What if the computer never gets anything to sink its tactical > teeth into, not because the human locked up the position, but because he > played PERFECTLY? As humans don't play perfectly this is an uninteresting question. > > I've already scored wins over those 3000-rated supermonsters on ICC, > including one against a machine that had over 7,000 wins and 36 losses. I > smashed it on the White side of a Sicilian Dragon at one-minute time > controls. Granted, they win most of the time, but you have to start > somewhere, and many players, even GMs, can't say they've beaten the machine > I beat, even once. You are an expert in bullet chess, ok. As long as you didn't prove your points over the board in long time controls, I don't take any of your claims. >> If it >> really comes down to "Calculate, calculate"? So, the way computers get >> their moves is absolutely not secondary. > > Computers actually use positional factors to evaluate the result of their > "calculation." Those positional factors are very shallow. That was one of the outcomes of the success of recent programs. It's all calculation and intelligent pruning. Positional factors are minor. > You think they find moves like Kh1 after castling through > brute tactical force? Sure.If They can calculate 10, 15 moves ahead, they might see, that this might be an advantage. No sophisticated knowledge is required. > >> Computers might help us in several insights, but how far >> this will go is only speculation. > > Only one way to find out... Yes, but it's still speculation. That's my point. > Weak player? I trained for four years and didn't chase rating points. My > average rate of improvement -- then and now -- is a very steady 100-150 elo > points per year. You still have to prove this. > I gather you are a very high-rated player, No, I am not. I don't make your disrespect for GMs and high ratings my own, but I wouldn't regard my number as high, and definitely not as very high. It's ok, that's all. Greetings, Ralf
|
| | | |
Date: 27 Apr 2006 21:24:42
From: Ray Gordon
Subject: Re: There is PLENTY of room at the top in chess
|
>> Yet the only way to refute my claim > > I don't have to. YOU made the claim, YOU have to prove it. As long as you > don't, I might say, it is just a claim, but nothing is proven. I didn't say "prove," I said "refute." "refutation" means "negative proof." >> So being booked up perfectly to move 30 is "here and there?" > > People are still not perfect to move 30. Maybe not YOUR people.... >And a guy like Aronyan shows, that you don't have to prepare your openings >like hell to go to 2700. Just imagine if he were booked up to move 30 with his game! He'd be 2950! That "room at the top" of which I speak. >>> Humans will never be able to calculate variations with the >>> perfection of computers. >> >> That sound you hear is Mikhail Tal laughing from his grave. > > Indeed, he was successful without perfect calculations. With his typical > style of play, Tal wouldn't have too big chances against Fritz. Nor would Tal have to play Fritz, and you don't know how much he would absorb from the machines. Videogamers are trained to beat machines at games. Chessplayers, apparently, are not. >> What computers do is simple: they play main-line openings until about >> move 10-12, then they look for a line that isn't well studied, play it to >> take the human out of book, and crush it tactically. >> >> You think a human can't adopt a similar approach? > > No. Simply because they can not play tactically on this level. Says who? Are you saying there are no possible shortcuts to extending one's combinative vision beyond that of a computer? We don't have to calculate every variation the way they do, but we certainly can improve that area sufficiently to neutralize the computer's *only* advantage over us. >> Most top GMs would probably tell you that there will come a day when a >> human can play perfect chess, or very close to perfect. > > Which GM said this for example? It was one of the top GMs who spoke of the possibility that humans would reclaim the throne. Might have been Adams before the Hydra match or Kasparov in an interview prior to his retirement. I'm sure if you ask the GMs (like at the World Open), they'd say pretty much the same thing. >> Before computers, it was believed that this was the case, and that we >> were already there (the "draw death" for example). > > Until when did most people believe this? Capablanca said so in the 30s. 40 > years later this was still not the case. The first K-K match had sixteen straight draws. Bad enough? It was thought that if either player had major weaknesses, that the other would exploit it. That proved to be false as well. >> Many, many times, usually with the person doing it breaking out in >> laughter at all the punctuation after the various moves in the >> annotation. Let's just say that Fischer's 60 memorable games aren't so >> memorable anymore. > > Whoever says something like that doesn't know too much about chess. Or knows a lot more.... Very few chessgames should survive being ripped apart by thirty-four years of history. I built my repertoire by copying Fischer's, but after about two years, I saw too many weaknesses in what he played and gradually drifted away from it. Now I wouldn't even consider most of the stuff he played, like that Benoni crap. >> Of course they are. Every move has a reason behind it, and that reason >> can be reverse-engineered. I learn something new every time I watch a >> computer play. > > Yes, but if a concrete situation deserves concrete calculations reverse > engineering will only help you in the post mortem analysis. Where the pattern recognition will help me in the next game. There is never anything wrong with seeing a better move being played. Great players often adopt the techniques with which they are beaten. It may be computer calculation, but it is still an exploitation of human miscalculation, so the human knows what he missed that the computer did not. >>> We will never be able to evaluate millions of positions per >>> second. >> >> We will never have to. We have the ability to create shortcuts that the >> computers do not. What the computer sees tactically, we can figure out >> positionally. > > It is an unproven claim that this is possible in a way much beyond what is > possible for humans today. All theory is unproven at first. I do have to have a style and a training regimen. I am sharing what I can of mine without divulging too many secrets. >> In this particular case, putting White's pawn at g3 is well worth the >> loss of two moves. Now WHY is this the case? There is a general >> principle here that explains this, and which also explains that time is >> the most overrated factor in the opening. > > I think there are older examples where one side deliberately gives away a > tempo in order to force the other side to make unfavorable pawn moves or > to put pieces on unfortunate positions. So this is not such a breathtaking > knowledge unheard of before the use of computers. Then why is 2...Qh5 not the main line? I also said that time is extremely overrated in the opening, which is more than just saying it's good "sometimes." I can't tell you how many times I bust a line up with Kh1 or Kh8 when the books want me to launch a premature attack instead. Same thing for Ng8-e7-c6, "trapping" the knight at b8, but also fortifying the scheme. >>> And this ability is the corner stone of the strenght of Fritz & >>> Co. These monsters have shown us, that concrete calculation beats >>> knowledge. >> >> Or they have shown us that inferior play in the opening is punished >> tactically. What if the computer never gets anything to sink its >> tactical teeth into, not because the human locked up the position, but >> because he played PERFECTLY? > > As humans don't play perfectly this is an uninteresting question. Which humans are you talking about? It is possible to play a perfect opening to X number of moves. I can usually do it to about move 17 now. When I quit in 1991, I was at move 12. I expect to be around move 25-30 in a few years. >> I've already scored wins over those 3000-rated supermonsters on ICC, >> including one against a machine that had over 7,000 wins and 36 losses. >> I smashed it on the White side of a Sicilian Dragon at one-minute time >> controls. Granted, they win most of the time, but you have to start >> somewhere, and many players, even GMs, can't say they've beaten the >> machine I beat, even once. > > You are an expert in bullet chess, ok. As long as you didn't prove your > points over the board in long time controls, I don't take any of your > claims. I hold my bullet games to a "slow" standard. Since most of my games are over within thirty moves, I have plenty of time to play the opening; not that I need it. A slower time control just gives me a ton of time to use and abuse, and the ability to win a won game with a five-second increment. Nakamura is the world's best one-minute player, btw. I'm only in the top 600 or so from the strongest servers. >>> If it >>> really comes down to "Calculate, calculate"? So, the way computers get >>> their moves is absolutely not secondary. >> >> Computers actually use positional factors to evaluate the result of their >> "calculation." > > Those positional factors are very shallow. That was one of the outcomes of > the success of recent programs. It's all calculation and intelligent > pruning. Positional factors are minor. The end analysis is always based on positional concepts, otherwise it would just be a material evaluation. >> You think they find moves like Kh1 after castling through brute tactical >> force? > > Sure.If They can calculate 10, 15 moves ahead, they might see, that this > might be an advantage. No sophisticated knowledge is required. They sure aren't finding checkmate with a move like that. They are evaluating the position at the end of the horizon to accomplish this. >>> Computers might help us in several insights, but how far >>> this will go is only speculation. >> >> Only one way to find out... > > Yes, but it's still speculation. That's my point. And only one way to find out, which is the point of my training. At least I have a plan and a purpose. >> Weak player? I trained for four years and didn't chase rating points. >> My average rate of improvement -- then and now -- is a very steady >> 100-150 elo points per year. > > You still have to prove this. Only to myself. My previous rating improvement is well-documented if USCF didn't lose the pre-1992 records. I even had a game published in the August 1990 Chess Life (Larry Evans' column, reader game of the month). Parker-Volovich (he was 2508 I was 1956). Crushed him out of the opening (missed a win with a sacrifice on f7), and lost on time while still winning. >> I gather you are a very high-rated player, > > No, I am not. I don't make your disrespect for GMs and high ratings my > own, but I wouldn't regard my number as high, and definitely not as very > high. It's ok, that's all. The rating you have is important to any debate; in this case, important for you to omit. If you are a weaker player than me, you won't understand how I train. If you are stronger, then you'll understand, but you'll just disagree. -- "Google maintains the USENET." -- The Honorable R. Barclay Surrick, Eastern District of PA Judge From Parker v. Google, E.D.Pa. #04-cv-3918
|
| | | | |
Date: 28 Apr 2006 10:17:55
From: Ralf Callenberg
Subject: Re: There is PLENTY of room at the top in chess
|
28.04.2006 03:24, Ray Gordon: >>> So being booked up perfectly to move 30 is "here and there?" >> People are still not perfect to move 30. > > Maybe not YOUR people.... Who is "YOUR people"? Those from planet earth? >> And a guy like Aronyan shows, that you don't have to prepare your openings >> like hell to go to 2700. > > Just imagine if he were booked up to move 30 with his game! > > He'd be 2950! Just a wild guess, not more. > >>> You think a human can't adopt a similar approach? >> No. Simply because they can not play tactically on this level. > > Says who? Even Kasparov, one of the strongest tactical players of all times, tried to avoid too tactical play against Deep Blue, because he knew, that his usual style of playing would not work against the machine. And Fritz and other current programs are even stronger now than Deep Blue was. It is simply impossible for humans to calculate to the depth as current top programs do. Added to this, humans make errors. They forget about pieces or put pieces on squares where they no longer are standing, simply oversee variations. If you calculate longer variations, the probability for such errors grows very fast. > > Are you saying there are no possible shortcuts to extending one's > combinative vision beyond that of a computer? At least up to now there is no hint into this direction. On the combinatorical side, the computers' advantage is still increasing. > We don't have to calculate every variation the way they do, but we certainly > can improve that area sufficiently to neutralize the computer's *only* > advantage over us. This is just an assumption. There is no hint, that this might be the case. >>> Most top GMs would probably tell you that there will come a day when a >>> human can play perfect chess, or very close to perfect. >> Which GM said this for example? > > It was one of the top GMs who spoke of the possibility that humans would > reclaim the throne. Might have been Adams before the Hydra match Aha... >> Until when did most people believe this? Capablanca said so in the 30s. 40 >> years later this was still not the case. > > The first K-K match had sixteen straight draws. Bad enough? Those draws had other reasons. They didn't appear because both sides had reached perfection. >>> Let's just say that Fischer's 60 memorable games aren't so >>> memorable anymore. >> Whoever says something like that doesn't know too much about chess. > > Or knows a lot more.... Which GM said so about this book? >> Yes, but if a concrete situation deserves concrete calculations reverse >> engineering will only help you in the post mortem analysis. > > Where the pattern recognition will help me in the next game. You always repeat this hope in pattern recognition, but this is just an assumption, a guess. You know how pattern recognition works? It relies on the ability to make fuzzy matches, to recognize a pattern, even if it slightly different from what you have seen before. Now, in a chess position a slight difference, a pawn on another square can make a huge difference. Pattern recognition helps you to find ideas, but at the end you have to make the calculation. >> It is an unproven claim that this is possible in a way much beyond what is >> possible for humans today. > > All theory is unproven at first. Well, but at least there is some evidence in the first place. Could you at least show this? > I also said that time is extremely overrated in the opening, which is more > than just saying it's good "sometimes." "Overrated" by whom? It is common knowledge, that if the position isn't about to get open tempi are not so important. They are important in sharp games, not so much in declined queen's gambit. >> As humans don't play perfectly this is an uninteresting question. > > Which humans are you talking about? The living ones. > > It is possible to play a perfect opening to X number of moves. I can > usually do it to about move 17 now. When I quit in 1991, I was at move 12. > I expect to be around move 25-30 in a few years. You claim you are perfect. Well, I think no other GM has done this before. Congratulations. Seriously: as long as you didn't prove this in a real tournament (no blitz or bullet crap), against strong players, I simply don't believe you. >> You are an expert in bullet chess, ok. As long as you didn't prove your >> points over the board in long time controls, I don't take any of your >> claims. > > I hold my bullet games to a "slow" standard. Since most of my games are > over within thirty moves, I have plenty of time to play the opening; not > that I need it. > > A slower time control just gives me a ton of time to use and abuse, and the > ability to win a won game with a five-second increment. Well, I can only repeat: play in a tournament and show it. Otherwise it is simply a claim, not more. >> Those positional factors are very shallow. That was one of the outcomes of >> the success of recent programs. It's all calculation and intelligent >> pruning. Positional factors are minor. > > The end analysis is always based on positional concepts, otherwise it would > just be a material evaluation. Sure, but there are only few concepts compared to what a GM knows. And a GM can find a move even without calculation, just based on positional knowledge. Computers are not able to do so. They have to calculate. >> Sure.If They can calculate 10, 15 moves ahead, they might see, that this >> might be an advantage. No sophisticated knowledge is required. > > They sure aren't finding checkmate with a move like that. They are > evaluating the position at the end of the horizon to accomplish this. Yes, but a GM could make such a move based on general assumptions. He doesn't have to calculate 10 or 15 moves. Again: a computer has only very small knowledge about positions. Only if you combine this with his ability to calculate a lot of variations, you get the strength we can see. If you would take the existing evaluations and stop the program thinking beyond move 3, you are back to the 80s with programs just good enough for amateurs. >> You still have to prove this. > > Only to myself. You make public claims, you are talking as if this all would be a fact. For me, all your claims about your increase in strength is just the talking of a mediocre chess player who spent a lot of time to become very good in bullet chess. > I even had a game published in the August 1990 Chess Life (Larry Evans' > column, reader game of the month). Parker-Volovich (he was 2508 I was > 1956). Crushed him out of the opening (missed a win with a sacrifice on > f7), and lost on time while still winning. It's not unheard of, that mediocre players have their big moments. That doesn't show anything. > > >>> I gather you are a very high-rated player, >> No, I am not. I don't make your disrespect for GMs and high ratings my >> own, but I wouldn't regard my number as high, and definitely not as very >> high. It's ok, that's all. > > The rating you have is important to any debate; in this case, important for > you to omit. I don't get it. First: I didn't say anything about the importance, I just said, my rating is not very high. Besides that I don't think, that the rating is actually so important in a debate. > If you are a weaker player than me, you won't understand how I train. If > you are stronger, then you'll understand, but you'll just disagree. My rating doesn't tell you whether I am weaker or stronger as you are. Greetings, Ralf
|
| | | | | |
Date: 28 Apr 2006 09:48:43
From: Ray Gordon
Subject: Re: There is PLENTY of room at the top in chess
|
>>>> So being booked up perfectly to move 30 is "here and there?" >>> People are still not perfect to move 30. >> >> Maybe not YOUR people.... > > Who is "YOUR people"? Those from planet earth? Those who aren't booked up to move 30. There are players who are. >>> And a guy like Aronyan shows, that you don't have to prepare your >>> openings like hell to go to 2700. >> >> Just imagine if he were booked up to move 30 with his game! >> >> He'd be 2950! > > Just a wild guess, not more. Well he'd be a lot stronger. >>>> You think a human can't adopt a similar approach? >>> No. Simply because they can not play tactically on this level. >> >> Says who? > > Even Kasparov, one of the strongest tactical players of all times, tried > to avoid too tactical play against Deep Blue, because he knew, that his > usual style of playing would not work against the machine. And Fritz and > other current programs are even stronger now than Deep Blue was. It is > simply impossible for humans to calculate to the depth as current top > programs do. Added to this, humans make errors. Only when they don't know what they are doing. >They forget about pieces or put pieces on squares where they no longer are >standing, simply oversee variations. If you calculate longer variations, >the probability for such errors grows very fast. For weaker players, it would. I didn't say it was easy to duplicate a computer's performance at chess, but it is definitely possible to do this or even to exceed it. I liken it to tial arts: no matter what punches and kicks you can land, your opponent can always come up with a way to attack you. There is no "brute force" way to win a chessgame against a sufficiently skilled opponent. The pre-computer era's chessplayers were weak because they could get away with it. These days, they can't. >> Are you saying there are no possible shortcuts to extending one's >> combinative vision beyond that of a computer? > > At least up to now there is no hint into this direction. On the > combinatorical side, the computers' advantage is still increasing. Over the *average* human. >> We don't have to calculate every variation the way they do, but we >> certainly can improve that area sufficiently to neutralize the computer's >> *only* advantage over us. > > This is just an assumption. There is no hint, that this might be the case. Computers don't have humans' "chess vision," especially in the opening. Show me a machine that won't play dxc5 if Black leaves it "hanging" in a queen-pawn game. >>>> Most top GMs would probably tell you that there will come a day when a >>>> human can play perfect chess, or very close to perfect. >>> Which GM said this for example? >> >> It was one of the top GMs who spoke of the possibility that humans would >> reclaim the throne. Might have been Adams before the Hydra match > > Aha... Choo. >>> Until when did most people believe this? Capablanca said so in the 30s. >>> 40 years later this was still not the case. >> >> The first K-K match had sixteen straight draws. Bad enough? > > Those draws had other reasons. They didn't appear because both sides had > reached perfection. The "other reasons" were based on the idea that getting a win for either player was "extremely difficult" due to the small amount of mistakes. It's two sides to the same coin. >>>> Let's just say that Fischer's 60 memorable games aren't so memorable >>>> anymore. >>> Whoever says something like that doesn't know too much about chess. >> >> Or knows a lot more.... > > Which GM said so about this book? They didn't have to. It comes out in the course of analysis of other games when they note the improvements. The games may be memorable for the fans, but they are no longer so relevant to chess theory. One game that has withstood the test of time in a way is Capablanca-shall, the "original shall attack," which became the main line and which is still the second most popular line. >>> Yes, but if a concrete situation deserves concrete calculations reverse >>> engineering will only help you in the post mortem analysis. >> >> Where the pattern recognition will help me in the next game. > > You always repeat this hope in pattern recognition, but this is just an > assumption, a guess. You know how pattern recognition works? It relies on > the ability to make fuzzy matches, to recognize a pattern, even if it > slightly different from what you have seen before. Now, in a chess > position a slight difference, a pawn on another square can make a huge > difference. Pattern recognition helps you to find ideas, but at the end > you have to make the calculation. Obviously. Kind of like if you have a system for picking stocks that is profitable, you still need to be able to analyze the financial jungle for things your stats don't pick up. Still, pattern recognition is the heart of chess improvement. If I didn't know how to improve my chessgame, my rating would never have hit 2000 back in the 1980s. >>> It is an unproven claim that this is possible in a way much beyond what >>> is possible for humans today. >> >> All theory is unproven at first. > > Well, but at least there is some evidence in the first place. Could you at > least show this? The priy evidence is that computers aren't spitting out perfect moves or anything close, and that a lot of the time they have the wrong side winning the game. That means it's *possible* for a human to exploit this. >> I also said that time is extremely overrated in the opening, which is >> more than just saying it's good "sometimes." > > "Overrated" by whom? It is common knowledge, that if the position isn't > about to get open tempi are not so important. They are important in sharp > games, not so much in declined queen's gambit. Even in "open" positions, time is still the most overrated factor. >>> As humans don't play perfectly this is an uninteresting question. >> >> Which humans are you talking about? > > The living ones. The original statement referred to perfect opening play. >> It is possible to play a perfect opening to X number of moves. I can >> usually do it to about move 17 now. When I quit in 1991, I was at move >> 12. I expect to be around move 25-30 in a few years. > > You claim you are perfect. Well, I think no other GM has done this before. > Congratulations. Seriously: as long as you didn't prove this in a real > tournament (no blitz or bullet crap), against strong players, I simply > don't believe you. I never said I was perfect for the whole game, just that I don't make many mistakes in the opening. Computers have made that even easier to accomplish now. >>> You are an expert in bullet chess, ok. As long as you didn't prove your >>> points over the board in long time controls, I don't take any of your >>> claims. >> >> I hold my bullet games to a "slow" standard. Since most of my games are >> over within thirty moves, I have plenty of time to play the opening; not >> that I need it. >> >> A slower time control just gives me a ton of time to use and abuse, and >> the ability to win a won game with a five-second increment. > > Well, I can only repeat: play in a tournament and show it. Otherwise it is > simply a claim, not more. Want to pay my entry fee and compensate me for missing two days of work? The reason I don't play now is that it's a big work in progress, and people only judge based on what they see now. They don't really judge players on which one is most likely to improve the most in the future, except when they try to link that to age. I could come back, get a 2300 performance rating, and people would see that as 500 points below where I need to go. I could be right on schedule to get there, but it wouldn't matter. For that privilege, I could waste time and entry fee, instead of staying home and working on my game. >>> Those positional factors are very shallow. That was one of the outcomes >>> of the success of recent programs. It's all calculation and intelligent >>> pruning. Positional factors are minor. >> >> The end analysis is always based on positional concepts, otherwise it >> would just be a material evaluation. > > Sure, but there are only few concepts compared to what a GM knows. And a > GM can find a move even without calculation, just based on positional > knowledge. Which is how it's possible to "outcalculate" a computer. >Computers are not able to do so. They have to calculate. So do humans. >>> Sure.If They can calculate 10, 15 moves ahead, they might see, that this >>> might be an advantage. No sophisticated knowledge is required. >> >> They sure aren't finding checkmate with a move like that. They are >> evaluating the position at the end of the horizon to accomplish this. > > Yes, but a GM could make such a move based on general assumptions. He > doesn't have to calculate 10 or 15 moves. How have the "general assumptions" performed so far for players like Adams against Hydra? >Again: a computer has only very small knowledge about positions. Only if >you combine this with his ability to calculate a lot of variations, you get >the strength we can see. If you would take the existing evaluations and >stop the program thinking beyond move 3, you are back to the 80s with >programs just good enough for amateurs. In the 1980s, Deep Thought was rated 2265. >>> You still have to prove this. >> >> Only to myself. > > You make public claims, you are talking as if this all would be a fact. > For me, all your claims about your increase in strength is just the > talking of a mediocre chess player who spent a lot of time to become very > good in bullet chess. I make my claims to present my point of view. >> I even had a game published in the August 1990 Chess Life (Larry Evans' >> column, reader game of the month). Parker-Volovich (he was 2508 I was >> 1956). Crushed him out of the opening (missed a win with a sacrifice on >> f7), and lost on time while still winning. > > It's not unheard of, that mediocre players have their big moments. That > doesn't show anything. 1956 is "mediocre?" I've also drawn Asa Hoffman (2500+) in a quad where we tied for first place. Try taking $12.50 out of Asa's pocket over the board. I also coached a high school with a 1400 player and four beginners from an 0-3 k midseason to a 5-5 finish and third place in the city league here by drilling them in the opening for three hours a day after school. >>>> I gather you are a very high-rated player, >>> No, I am not. I don't make your disrespect for GMs and high ratings my >>> own, but I wouldn't regard my number as high, and definitely not as very >>> high. It's ok, that's all. >> >> The rating you have is important to any debate; in this case, important >> for you to omit. > > I don't get it. First: I didn't say anything about the importance, I just > said, my rating is not very high. Besides that I don't think, that the > rating is actually so important in a debate. So I have to prove myself, but you can get away with already being weaker than me yet still with a better grasp of chess theory than me? I'm not a rating snob. I play people rated any level. Unless you have worked your way up through the rating classes that I have, you won't understand what one needs to do to continue improving at that level. It doesn't mean you aren't entitled to opinions on how to improve, but it does mean you won't understand what I'm doing the way you would if you had done it yourself. >> If you are a weaker player than me, you won't understand how I train. If >> you are stronger, then you'll understand, but you'll just disagree. > > My rating doesn't tell you whether I am weaker or stronger as you are. So you're proud of clouding what is supposed to be an open debate? My original post merely pointed out that GM play is still weak enough for a dedicated, talented player to come in and dominate. If we were discussing say tic-tac-toe, this wouldn't be the case, as the Mott Street Chicken still has never lost. -- "Google maintains the USENET." -- The Honorable R. Barclay Surrick, Eastern District of PA Judge From Parker v. Google, E.D.Pa. #04-cv-3918
|
| | | | | | |
Date: 28 Apr 2006 17:36:42
From: Earine
Subject: Re: There is PLENTY of room at the top in chess
|
Ray Gordon wrote: >>>>> So being booked up perfectly to move 30 is "here and there?" >>>> People are still not perfect to move 30. >>> Maybe not YOUR people.... >> Who is "YOUR people"? Those from planet earth? > > Those who aren't booked up to move 30. There are players who are. etc. etc. ad infinitum.... Ray Gordon lives in his own fantasy world and has very strong beliefs that are against common knowledge in chess. He makes a lot of unfounded claims. And it is impossible to argue with him, as this thread and several others have shown. If nobody replied to him then perhaps he would notice that he is not getting any attention here and would go away. Why should we care about his nonsense? -Earine
|
| | | | | | | |
Date: 28 Apr 2006 17:47:05
From: Ray Gordon
Subject: Re: There is PLENTY of room at the top in chess
|
7 >>>>>> So being booked up perfectly to move 30 is "here and there?" >>>>> People are still not perfect to move 30. >>>> Maybe not YOUR people.... >>> Who is "YOUR people"? Those from planet earth? >> >> Those who aren't booked up to move 30. There are players who are. > > etc. etc. ad infinitum.... > > > Ray Gordon lives in his own fantasy world and has very strong beliefs that > are against common knowledge in chess. So did Steinitz. One has to have "delusions" to do anything meaningful in chess. > If nobody replied to him then perhaps he would notice that he is not > getting any attention here and would go away. Why should we care about his > nonsense? Since when did YOU become a WE? Talk about delusional. -- "Google maintains the USENET." -- The Honorable R. Barclay Surrick, Eastern District of PA Judge From Parker v. Google, E.D.Pa. #04-cv-3918
|
|
Date: 27 Apr 2006 06:47:24
From: Ralf Callenberg
Subject: Re: There is PLENTY of room at the top in chess
|
Ray Gordon schrieb: > My point was that the gap is large enough for a human to easily enter it > without even approaching what the computers can do. And my point is, that this is taken out of thin air. There is also a huge gap between Michael Johnson and a cheetah - nevertheless humans are simply not able to enter this gap much deeper. > When Fischer was at his peak, there were people who would have thought that > he was closer to perfection than we think the computers are now. He won > twenty straight games against top competition. > > Look at what Topalov did to the field last year if you want to see what's > possible with a strong human player. He was clearly dominant. Yes, this happened again and again,. Lasker and Capablanca are examples of the past. There will always be one or two dominating their era. And what does this tell us? Probably the strength has increased because knowledge has increased, the use of improved techniques. But by which gin the strength has increased is a matter of speculation. And the advance of machines doesn't tell us anything about the possibities of the human mind. They are simply not related. So your statement "If Fritz can do it, humans can do it" is just a claim without the slightest trace of proof for it. Computers and humans play chess completely different. As a Eurofighter flies completely different than a falcon. Greetings, Ralf
|
| |
Date: 27 Apr 2006 10:27:52
From: Ray Gordon
Subject: Re: There is PLENTY of room at the top in chess
|
>> My point was that the gap is large enough for a human to easily enter it >> without even approaching what the computers can do. > > And my point is, that this is taken out of thin air. There is also a > huge gap between Michael Johnson and a cheetah - nevertheless humans > are simply not able to enter this gap much deeper. Humans can learn chess from computers, since the computers have to divulge the moves they play, if not the reasons behind the move. Knowing how a cheetah runs so fast will not help a human to run faster, but learning that your pet opening has a hidden move that gives you a +3.43 evaluation a few moves down the road definitely will. Your example is not relevant to this issue. If the gap were "not very deep" then computers wouldn't be crushing our GMs because our GMs would already be close to perfect. The fact remains, if Adams can get slaughtered by Hydra in 25-30 moves each game, then Adams is a pretty damn weak player from an objective standpoint. >> When Fischer was at his peak, there were people who would have thought >> that >> he was closer to perfection than we think the computers are now. He won >> twenty straight games against top competition. >> >> Look at what Topalov did to the field last year if you want to see what's >> possible with a strong human player. He was clearly dominant. > > Yes, this happened again and again,. Lasker and Capablanca are examples > of the past. There will always be one or two dominating their era. And > what does this tell us? That it's possible for someone to dominate his peers in chess, and that the dreaded "draw death" isn't anywhere close on the horizon. > Probably the strength has increased because knowledge has increased, In large part because of computers! > the use of improved techniques. But by which gin the strength has > increased is a matter of speculation. Not really, since we can run the games of every generation through the silicon to see who knew what. >And the advance of machines > doesn't tell us anything about the possibities of the human mind. Of course it does! It shows us the ultimate perfection that can be achieved when humans reduce chess to a set of general principles that allow them to never lose. >They > are simply not related. So your statement "If Fritz can do it, humans > can do it" is just a claim without the slightest trace of proof for it. Hardly: Fritz can easily teach us what is correct to do in a given position. We can learn openings, positional themes, tactics, etc. from it. The moves it plays are played for reasons, very sound reasons, and it's up to us to uncover them. Many players learn by studying GM games and then figuring out for themselves why the moves were played. With computers, we're doing the same thing, only with better moves to study. At least I *thought* this was what people were doing, because if they aren't, you better reserve a seat for me at the world championship sometime in the next decade since my training has me on a much faster track. > Computers and humans play chess completely different. As a Eurofighter > flies completely different than a falcon. How computers arrive at their moves is secondary to the fact that their moves are the ones that the top players need to be studying. My game has already changed immensely because of what computers taught me, and I spent 75-80 hours a week training back in the late 1980s, when they were not reliable training tools (strongest computer was about 2250 and only reliable in heavily tactical situations at slow time controls). Video game players learn to master video games without the computers that run them telling them how. They figure it out "over the board" or through analysis of why they "lost" their turns. In a videogame, if you make a mistake, you get killed, whereas against a chess computer, the same thing happens. -- "Google maintains the USENET." -- The Honorable R. Barclay Surrick, Eastern District of PA Judge From Parker v. Google, E.D.Pa. #04-cv-3918
|
|
Date: 27 Apr 2006 00:22:35
From: Ralf Callenberg
Subject: Re: There is PLENTY of room at the top in chess
|
26.04.2006 23:09, Ray Gordon: > Look at the top players against the top computers: they can't compete. > > What this means is that even our "Super GMs" are quite beatable. If Hydra > and Fritz can do it, so can a human. We just aren't as good at chess as we > like to think. When reading this, my first reaction was to look at the header to detect our friend from Yugoslavia, but no, this is indeed from Ray Gordon himself. So finally, we know, that birds can fly Mach 1.5 - if Super Hornet and Eurofighter can do it, so can a bird. Our feathered friends just aren't as good at flying as they like to think. Greetings, Ralf
|
| |
Date: 27 Apr 2006 01:08:42
From: Ray Gordon
Subject: Re: There is PLENTY of room at the top in chess
|
>> Look at the top players against the top computers: they can't compete. >> >> What this means is that even our "Super GMs" are quite beatable. If >> Hydra and Fritz can do it, so can a human. We just aren't as good at >> chess as we like to think. > > > So finally, we know, that birds can fly Mach 1.5 - if Super Hornet and > Eurofighter can do it, so can a bird. Our feathered friends just aren't as > good at flying as they like to think. In 1972, if someone had talked about a "supercomputer" that could play perfect chess, what would the public have thought of Fischer's chances against it? We sort of have an answer to that question because when the debate began in earnest -- after Deep Thought broke 2200 -- it was thought that computers would "hit the wall" against the GMs, that the GMs could "control the game" and force draw after draw, if not exploit the machine's weaknesses and win. It was thought at the time that GMs made few mistakes, and that even a perfect machine would have a difficult time exploiting them. We now know that this is not even close to the case. One does not need to match the strength of the computers to surpass the top human players, and the point of my post was to note that there is an awful lot of room between our top players and "perfect" play, room that can be exploited by another human who does not even have to come close to the machines to be far superior to even the best player in the world. -- "Google maintains the USENET." -- The Honorable R. Barclay Surrick, Eastern District of PA Judge From Parker v. Google, E.D.Pa. #04-cv-3918
|
| | |
Date: 28 Apr 2006 12:47:15
From: Kenneth Sloan
Subject: Re: There is PLENTY of room at the top in chess
|
Earine <[email protected] > writes: > ... > If nobody replied to him then perhaps he would notice that he is not > getting any attention here and would go away. good idea - you go first. -- Kenneth Sloan [email protected] Computer and Information Sciences (205) 934-2213 University of Alabama at Birmingham FAX (205) 934-5473 Birmingham, AL 35294-1170 http://www.cis.uab.edu/sloan/
|
| | |
Date: 27 Apr 2006 10:40:16
From: Ralf Callenberg
Subject: Re: There is PLENTY of room at the top in chess
|
27.04.2006 07:08, Ray Gordon: > In 1972, if someone had talked about a "supercomputer" that could play > perfect chess, what would the public have thought of Fischer's chances > against it? This only proves, that they underestimated the technical possibilities and does not say anything about your claim. When Lilienthal started his first attempts to fly, nobody would have thought, that a 100 years later planes could fly as fast as the sound. > We sort of have an answer to that question because when the debate began in > earnest -- after Deep Thought broke 2200 -- it was thought that computers > would "hit the wall" against the GMs, that the GMs could "control the game" > and force draw after draw, if not exploit the machine's weaknesses and win. > It was thought at the time that GMs made few mistakes, and that even a > perfect machine would have a difficult time exploiting them. At least I never heard this claim. What I read was, that the computers would hit a wall because of the exploding numbers of possibilities and that they would never gain the knowledge a human has to circumvent this limitation. Well, they have been right about the second half: computers have only low level of chess knowledge as humans use it, but they underestimated the effectiveness of pruning strategies in combination with the increase in calculating power. That even a perfect machine would have difficulties winning against a GM is something which doesn't sound plausible so that I strongly doubt, that a lot of people shared this opinion. > One does not need to match the strength of the computers to surpass the top > human players, and the point of my post was to note that there is an awful > lot of room between our top players and "perfect" play, Yes, as there is also a lot of room between the current world class sprinters and a cheetah, nevertheless the the world records increase only in tiny steps since many years. > room that can be > exploited by another human who does not even have to come close to the > machines to be far superior to even the best player in the world. The machines have shown, that GMs are not nearly perfect. OK. But that humans have so much room above the GMs does simply not follow from this. It could be, that the human brain simply is reaching a wall it can not penetrate. I don't say, that this is the case, but your machine-argument doesn't show anything in this direction. Greetings, Ralf
|
| |
Date: 27 Apr 2006 00:59:58
From: Bjoern
Subject: Re: There is PLENTY of room at the top in chess
|
Ralf Callenberg wrote: > 26.04.2006 23:09, Ray Gordon: > >> Look at the top players against the top computers: they can't compete. >> >> What this means is that even our "Super GMs" are quite beatable. If >> Hydra and Fritz can do it, so can a human. We just aren't as good at >> chess as we like to think. > > When reading this, my first reaction was to look at the header to detect > our friend from Yugoslavia, but no, this is indeed from Ray Gordon himself. > > So finally, we know, that birds can fly Mach 1.5 - if Super Hornet and > Eurofighter can do it, so can a bird. Our feathered friends just aren't > as good at flying as they like to think. Obviously there's also a plenty of room at the top of athletics, I mean, 9.x seconds for 100m? Just look at motorbikes and racing cars. Or throwing a javelin to 100m, Roman/medieval ballistas already managed 300+m! If motorbikes and ballistas can do that, so can humans! Quite simply humans are just cannot play chess in the same way computers do, because despite increasingly reasonable positional evaluation algorithms the playing strength of computers is still foremost based on their ability to check hundreds of thousands/millions of options in seconds (=brute-force). Well, Ray, you would be more amusing if you'd come up with something new (unless that was actually you annoucing that Caro-Kann book, I'm sure that would be quite funny). Bj�rn
|
|