How deep does a human think when playing chess?

[metax]

Hi Nick,
I have to agree with everyone saying that 6-7 plys isn’t very deep for a human or a computer. Also computers use selective search, why would you want to calculate a line that isn’t going to matter it would be pointless.
About your other question regarding how long it’ll be before programs can play as well as humans at 6 ply deep. I think most GM’s would lose to rybka if they were only allowed to look 6 ply deep. A GM would have to look at some very deep positional advantages to beat rybka at 6 ply. Vas rates rybka 3 around 2400 at 5 ply deep.

As Rybka is probably reporting false search depths, the depth which is reported as 5 ply by Rybka could be much higher.

[david_32626]

Hi Nick,
I have to agree with everyone saying that 6-7 plys isn’t very deep for a human or a computer. Also computers use selective search, why would you want to calculate a line that isn’t going to matter it would be pointless.
About your other question regarding how long it’ll be before programs can play as well as humans at 6 ply deep. I think most GM’s would lose to rybka if they were only allowed to look 6 ply deep. A GM would have to look at some very deep positional advantages to beat rybka at 6 ply. Vas rates rybka 3 around 2400 at 5 ply deep.

As Rybka is probably reporting false search depths, the depth which is reported as 5 ply by Rybka could be much higher.

What makes you say that?

[slobo]

Hi; Some years ago I read that Capablanca was asked how many
moves he thinks ahead. Unfortunately I don't remember the exact
number, but it was surprisingly low, maybe 4 or 5.

No, no, it was much more than 5, I think 7, 9, or even 10 in some lines. One day, however, Capablanca lost a game against an opponent not so impressive, an almost unknowen fellow, and the journalists asked that guy, how many moves did he see in advance, while he played against Capablanca. The guy answered:
"I see only one move ahead... but it is the best one!"

I liked that one very much.

[Spacious_Mind]

Where exactly did you get this article?
Thanks

Hi Marek,

It comes from the User Manual for Chess 7.0 written by Larry Atkin. Peter Frey was (don't know if he still is) a professor at Northwestern University teaching courses in Psychology and Computer Science. Editor and author of and contributor to the definitive text on computer chess (back in his day) CHESS SKILL IN MAN AND MACHINE.

He worked very closely together with Larry Atkin and David Slate and I am sure that you would know their World Championship successes with the Program CHESS (4.0, 4.6 etc). These were later ported to Home PC and improved upon as CHESS versions 7.0 and 7.5 distributed by ODESTA.

I don't know how you can refute what is written when he specifically writes about DeGroots research that was made over many years. It is crazy to think that flippant assumptions of being better should be made without providing counterproof research that refutes DeGroot. It clearly states in the text I gave you that further research was made in the US which confirmed his findings.

If you read my first questions, I specifically asked if this was ever refuted... and I have heard nothing sofar here which indicates that it was refuted. Read his observations again and you will see that they make total sense.

Again my comment about David Levy making a mistake in his bet. Think about it. If the bet had have been made (in the Hey Day of early chess computers where he made World News! with his claim) that he would not be beaten in 10 years by a computer. Would it not perhaps have been better for him to have quoted DeGroot's research and stated he would not be beaten by a computer that plays like a human at 6 or 7 ply? Why ? because everyone went forward trying to get better with pure raw speed. What if the focus back then would have been to take up his challenge... where would computers be today at full speed ...4000 ELO perhaps who knows... in the meantime Rybka at 5 Ply (2ply) is hardly better than Colossus 3 written in 1985... at the human game of chess...

I don't know why these thoughts should upset anyone... I am just expressing different thoughts and perhaps even future challenges.. I know one thing, the GM's today would probably still be playing competitions with engines had the original challenge been worded differently The sport would have been much bigger today also and a lot more interesting... for board chess players.

Here is the text where I took it from my Website.



http://www.spacious-mind.com/html/commo ... s_7_0.html

Best regards

Nick

[david_32626]

Where exactly did you get this article?
Thanks

Hi Marek,

It comes from the User Manual for Chess 7.0 written by Larry Atkin. Peter Frey was (don't know if he still is) a professor at Northwestern University teaching courses in Psychology and Computer Science. Editor and author of and contributor to the definitive text on computer chess (back in his day) CHESS SKILL IN MAN AND MACHINE.

He worked very closely together with Larry Atkin and David Slate and I am sure that you would know their World Championship successes with the Program CHESS (4.0, 4.6 etc). These were later ported to Home PC and improved upon as CHESS versions 7.0 and 7.5 distributed by ODESTA.

I don't know how you can refute what is written when he specifically writes about DeGroots research that was made over many years. It is crazy to think that flippant assumptions of being better should be made without providing counterproof research that refutes DeGroot. It clearly states in the text I gave you that further research was made in the US which confirmed his findings.

If you read my first questions, I specifically asked if this was ever refuted... and I have heard nothing sofar here which indicates that it was refuted. Read his observations again and you will see that they make total sense.

Again my comment about David Levy making a mistake in his bet. Think about it. If the bet had have been made (in the Hey Day of early chess computers where he made World News! with his claim) that he would not be beaten in 10 years by a computer. Would it not perhaps have been better for him to have quoted DeGroot's research and stated he would not be beaten by a computer that plays like a human at 6 or 7 ply? Why ? because everyone went forward trying to get better with pure raw speed. What if the focus back then would have been to take up his challenge... where would computers be today at full speed ...4000 ELO perhaps who knows... in the meantime Rybka at 5 Ply (2ply) is hardly better than Colossus 3 written in 1985... at the human game of chess...

I don't know why these thoughts should upset anyone... I am just expressing different thoughts and perhaps even future challenges.. I know one thing, the GM's today would probably still be playing competitions with engines had the original challenge been worded differently The sport would have been much bigger today also and a lot more interesting... for board chess players.

Here is the text where I took it from my Website.



http://www.spacious-mind.com/html/commo ... s_7_0.html

Best regards

Nick

When the common computer has as much processing power as the human brain then we might get to see a chess program that understands and plays chess more like a GM. Even though a computer looks at more positions during a game the human brain is processing much more information.

[Spacious_Mind]

When the common computer has as much processing power as the human brain then we might get to see a chess program that understands and plays chess more like a GM. Even though a computer looks at more positions during a game the human brain is processing much more information.

Hi David,
I am not sure if I would be willing to agree to that either. The quote was that a human looks at 35 future positions on average. (in say in 3 minutes (2hrs/40) ) What if you take Father's tactical comments into consideration and a computer also is restricted to looking only say 7 ply deep and 35 future positions, but someday it is programmed to be tactically better. Then perhaps a GM would be beaten at 6 or 7 ply right? maybe ?

It would then be just a matter of who is tactically better? What if a program can be taught within these confines to learn from its mistakes...what then?

If you assume that both human and computer search selectively then it would be just a matter of who selectively searches (tactically?) better and not deeper?

Best regards

Nick

[Uri Blass]

When the common computer has as much processing power as the human brain then we might get to see a chess program that understands and plays chess more like a GM. Even though a computer looks at more positions during a game the human brain is processing much more information.

Hi David,
I am not sure if I would be willing to agree to that either. The quote was that a human looks at 35 future positions on average. (in say in 3 minutes (2hrs/40) ) What if you take Father's tactical comments into consideration and a computer also is restricted to looking only say 7 ply deep and 35 future positions, but someday it is programmed to be tactically better. Then perhaps a GM would be beaten at 6 or 7 ply right? maybe ?

It would then be just a matter of who is tactically better? What if a program can be taught within these confines to learn from its mistakes...what then?

If you assume that both human and computer search selectively then it would be just a matter of who selectively searches (tactically?) better and not deeper?

Best regards

Nick

I believe that strong humans clearly look at more than 35 future positions on average in 3 minutes.

I do not trust humans to give correct details about their thinking process.

For example strong humans may not tell you that they considered Ra8+ in the following diagram because it is easily refuted by Bxa8 but it does not mean that they do not do it.

I expect every strong human to consider Ra8+ and reject it because of Bxa8 in less than a second.

[D]2k3r1/7p/2b5/8/8/R1B5/8/5K2 w - - 0 1

I believe that in part of the cases humans even do not know that they consider something so the problem is not only that humans may not be honest about their thinking process.

Humans do a lot of things that they do not know how they do it.

They may walk without knowing all the calculations that their brain does in order to walk and they may play chess moves without knowing all the calculation that their brain does in order to choose the chess moves.

Uri

[Spacious_Mind]

I believe that strong humans clearly look at more than 35 future positions on average in 3 minutes.

I do not trust humans to give correct details about their thinking process.

For example strong humans may not tell you that they considered Ra8+ in the following diagram because it is easily refuted by Bxa8 but it does not mean that they do not do it.

I expect every strong human to consider Ra8+ and reject it because of Bxa8 in less than a second.

[D]2k3r1/7p/2b5/8/8/R1B5/8/5K2 w - - 0 1

I believe that in part of the cases humans even do not know that they consider something so the problem is not only that humans may not be honest about their thinking process.

Humans do a lot of things that they do not know how they do it.

They may walk without knowing all the calculations that their brain does in order to walk and they may play chess moves without knowing all the calculation that their brain does in order to choose the chess moves.

Uri

I think in your position above a human chess player does not even spend any energy evaluating Ra8. Experience has tought him as a beginnner (child) after having moved to a8 one or two times in the past that it is a bad move. There is no consideration for a8 at all it is rejected the moment you see the position. You have already learned from similar experiences that it is wrong. No calculations needed by a human here, just a visual board inspection. But let an absolute beginner get to this position, even though he might try to calculate moves, he/she probably still would play the move at least one time in their absolute beginnings and learn it was a wrong move So both experienced human and computer reject this move in a fraction of a second ?

Best regards

Nick

ps.. in your example above your thoughts would be on just one thing and that is to capture the pawn (unless you see some unnatural move made by your opponent). You wouldn't even attempt to evaluate anything close to 35 positions.

regards

Nick

[Rolf]

I do not trust humans to give correct details about their thinking process.

I wouldnt trust them in any question that have anything to do with their chess. The naivety of de Groot is huge and that of his followers all the more. Fact is there is no observer independent research possible with the psychologist as observer. That already makes it clear.

For the given reasons I would never believe publications about not existing eidetics as factor number one. I mean this is what it's all about in professional chess. Without you must not even step on stage.

I was kind of surprised that in the longer book quote it was mentioned that the lack of special memory talents was allegedly shown in the years 1920, so shortly after the Soviet Revolution. But here it's clear that the fact as such would be treated like a secret. Even a couple of years ago I saw Karpov on video answering if he had special skills of memory that no, not at all. This is simply ridiculous. Of course we must see that you cant always explain the obvious to every new naive person who asks you. Perhaps he meant no in respect to the known assumption that chessplayer learn by heart all their openings which is certainly true and trivial all the same but this is not the preparation a professional is mainly doing. He doesnt memorize lines like a stupid computer or a patzer, but he gets some huge data material about certain issues, that is so huge that normal mortals like you and me wouldnt literally be incapable of learning this by heart. Most of us even have no clear vision of the board with its pieces. In that case memorizing means that we learn a line like a song but we dont see exactly the positions in advance. But this is exactly what professional players do all day long. One could say that they play hundreds of games during a single game of chess.

To seriously deny that this affords a clear memory eidetics is like denying that we must breathe. Because all this must happen under highest speed and how would you do it withoit eidetics. Because you must in reality do what de Groot thought he had rejected. Masters must work with exactest pictures of the concrete positions with the slightest details, otherwise they couldnt play at all. The fallacy in de Groot's project is that the masters knew all the positions de Groot could possibly create as a weaker player. So, more or less they played a game with him, because they knew that Dr Euwe was the Dutch candidate player for whom all the results could have been of interest. Why should they have told him exactly up to what heights their capabilities were going?

I dont know if you have an idea what eidetics means. The most ridiculous finding of de Groot was that average players are similar to professionals if the positions were scrumbled meaningless. But eidetics means that I can visualize even the most meaningless phonebooks of the world. Would I really have such a problem to locate the max. 32 pieces on a chessboard?

[Spacious_Mind]

I do not trust humans to give correct details about their thinking process.

I wouldnt trust them in any question that have anything to do with their chess. The naivety of de Groot is huge and that of his followers all the more. Fact is there is no observer independent research possible with the psychologist as observer. That already makes it clear.

For the given reasons I would never believe publications about not existing eidetics as factor number one. I mean this is what it's all about in professional chess. Without you must not even step on stage.

I was kind of surprised that in the longer book quote it was mentioned that the lack of special memory talents was allegedly shown in the years 1920, so shortly after the Soviet Revolution. But here it's clear that the fact as such would be treated like a secret. Even a couple of years ago I saw Karpov on video answering if he had special skills of memory that no, not at all. This is simply ridiculous. Of course we must see that you cant always explain the obvious to every new naive person who asks you. Perhaps he meant no in respect to the known assumption that chessplayer learn by heart all their openings which is certainly true and trivial all the same but this is not the preparation a professional is mainly doing. He doesnt memorize lines like a stupid computer or a patzer, but he gets some huge data material about certain issues, that is so huge that normal mortals like you and me wouldnt literally be incapable of learning this by heart. Most of us even have no clear vision of the board with its pieces. In that case memorizing means that we learn a line like a song but we dont see exactly the positions in advance. But this is exactly what professional players do all day long. One could say that they play hundreds of games during a single game of chess.

To seriously deny that this affords a clear memory eidetics is like denying that we must breathe. Because all this must happen under highest speed and how would you do it withoit eidetics. Because you must in reality do what de Groot thought he had rejected. Masters must work with exactest pictures of the concrete positions with the slightest details, otherwise they couldnt play at all. The fallacy in de Groot's project is that the masters knew all the positions de Groot could possibly create as a weaker player. So, more or less they played a game with him, because they knew that Dr Euwe was the Dutch candidate player for whom all the results could have been of interest. Why should they have told him exactly up to what heights their capabilities were going?

I dont know if you have an idea what eidetics means. The most ridiculous finding of de Groot was that average players are similar to professionals if the positions were scrumbled meaningless. But eidetics means that I can visualize even the most meaningless phonebooks of the world. Would I really have such a problem to locate the max. 32 pieces on a chessboard?

I do not trust humans to give correct details about their thinking process.

I wouldnt trust them in any question that have anything to do with their chess. The naivety of de Groot is huge and that of his followers all the more. Fact is there is no observer independent research possible with the psychologist as observer. That already makes it clear.

For the given reasons I would never believe publications about not existing eidetics as factor number one. I mean this is what it's all about in professional chess. Without you must not even step on stage.

I was kind of surprised that in the longer book quote it was mentioned that the lack of special memory talents was allegedly shown in the years 1920, so shortly after the Soviet Revolution. But here it's clear that the fact as such would be treated like a secret. Even a couple of years ago I saw Karpov on video answering if he had special skills of memory that no, not at all. This is simply ridiculous. Of course we must see that you cant always explain the obvious to every new naive person who asks you. Perhaps he meant no in respect to the known assumption that chessplayer learn by heart all their openings which is certainly true and trivial all the same but this is not the preparation a professional is mainly doing. He doesnt memorize lines like a stupid computer or a patzer, but he gets some huge data material about certain issues, that is so huge that normal mortals like you and me wouldnt literally be incapable of learning this by heart. Most of us even have no clear vision of the board with its pieces. In that case memorizing means that we learn a line like a song but we dont see exactly the positions in advance. But this is exactly what professional players do all day long. One could say that they play hundreds of games during a single game of chess.

To seriously deny that this affords a clear memory eidetics is like denying that we must breathe. Because all this must happen under highest speed and how would you do it withoit eidetics. Because you must in reality do what de Groot thought he had rejected. Masters must work with exactest pictures of the concrete positions with the slightest details, otherwise they couldnt play at all. The fallacy in de Groot's project is that the masters knew all the positions de Groot could possibly create as a weaker player. So, more or less they played a game with him, because they knew that Dr Euwe was the Dutch candidate player for whom all the results could have been of interest. Why should they have told him exactly up to what heights their capabilities were going?

I dont know if you have an idea what eidetics means. The most ridiculous finding of de Groot was that average players are similar to professionals if the positions were scrumbled meaningless. But eidetics means that I can visualize even the most meaningless phonebooks of the world. Would I really have such a problem to locate the max. 32 pieces on a chessboard?

Hi Rolf,

Have you researched this at all.. or are you just making opinions again... No one told anyone anything.. They were asked to do some tests... so you think people would withhold stuff in tests. Tests by nature are challenges and if you think you are good you would want to prove you are good in a test. Give a chess player a problem and he would want to solve it right? ... or do you think he would hold back ?

Since we are all knowing .. you Rolf have one second to solve this one j/k

[D]1n6/3q2B1/3R3Q/2k1N3/6p1/1nN3K1/4P3/8 w - - 0 1

(no cheating white to move)