I found it a bit ironic that at depth 18 vs depth 17 black suddenly has to give up the rook on a8. And also Rybka doesn't give anything else to explain why. Black box =/
Thanks KC, very interesting video (as are your other ones) and your point about the mainline syndrome is interesting. My view on it is that yes, computers are better than humans at chess but if we start taking the computer's suggestion as the "correct" move, then our thinking will become misguided because we can't explain where this move comes from. Humans do have the ability to be innovative and create and refute hypotheses, but chess is a closed enough system to yield to brute-force search.
i think its cheesy that you backed off engine analysis because of a GM's comments. An engine's algorithm finds the quickest a most effective way to 1) gain piece advantage, and 2) mate the opponent. To assume that you can take the second line as opposed to the main line and not be a Mainline Servant and still be in a better position is silly. The mainline is just that, the main line of moves that will accomplish the above to goals the quickest. running out room to type...
I think the point here is that eng analysis is more suited to analyzing games between engines. Engines miss the human aspect of analysis, and only best moves found by humans take this account into consideration.
So, the best move against an engine is not necessarily the best move against a human.
But I agree, the engine (these days) will be ultimately more correct even with subtle things like order of moves in a sequence. Engines are stronger ultimately than humans.
I'm really happy you tied zen and the art of motorcycle maintenance into this discussion! Engine analysis is a really thought-provoking application of Pirsig's ideas about subject-object interaction and the emergence of 'quality'.
Any chance you could post the video of that Short game as a reply? (btw, it's pretty cool that GMs are checking out the videos you do of their games, props man).
I also believe that you're kind of overstating the obvious here KC, no offense. Anyone and everyone not only playing chess, but even further, using the engine to analyze games is certainly aware of the fact that what would normally be considered even brilliant by human standards, is most often brushed away by the likes of Rybka and others in favor of the 1,855,907th Var. If the sole point is to say, make sure you balance your play and realize that your opponents are only human too, then kudos.
There is actually a deeper aspect about the lack of "internal quality" focus, which I will try and expand on in another video part to this. Without such a focus, it is as if we are just accepting the results of engines without question.
KC,where can I buy Rybka and learn how to make chess vids?
Can you consider becoming sort of People's Enlightenment Guru and spread the light of knowledge in philosophy,psycology,natural sciences,web administration,computers,chess on youtube?
Thanks for the reply KC. I agree that it;s very difficult NOT to just accept a Rybka-like response, after all, it IS A COMPUTER, RIGHT?! Tongue-in-cheek. I'll have to watch your second installment here and respond...
@JohnnySkillish Yes but it is like accepting an artefact of "external quality" without having access to its "internal quality". You don't understand that artefact -it is just a black box. So despite being technically great, it is also unevolvable, unexplainable, etc.
Great analogy. I did watch the second installment, which was fascinating once again. I would like to read your response to it my post on that page if you are so inclined. And anyone else's too for that matter. Thanks again! JS
I am always attempting to be cognoscente of the fact that when I use Rybka to analyze a game I've played, that BOTH sides are being analyzed and that the program is always going to be looking for equalization, which is why true sac's are never made and also why "simple" and "obvious" moves are often simply dispensed of by Rybka. Therefore, I insert other "human" moves based on my own thoughts in a position and see where it goes. Surprisingly, sometimes it actually ends up better in the end.
I think a lot of people noticed your mistake at the end of that game's commentary. I know I was thinking the same thing after watching the position. All the moves are made to get to a position where the bishop can be captured. I suppose the lesson is that there needs to be preparation when commenting on games to avoid running into claiming a computer zwischenzug as a legitimate brilliant or winning move.
I'm looking forward to future commentary, especially if you stick by what you've said here
Very interesting discussion! I hope that it will still be alive when I return (since I am too tired to participate as of now). I have done extensive work on the subject (as part of a larger study of AI in general) and I am really glad that you brought it up!
Very interesting dialogue there Tryfon, it reflects some things that i've said to friends before. I've always had the opinion that chess is about truth. Both players will see the position differently, have different ideas about what's going to happen next, and what's possible tactically, it's the person that sees the truth of the position better that is going to win.
The problem with engines, and society in general, is when people stop questioning the answers they're given. ;)
Very interesting indeed. I wonder if 'Mainline Syndrome' exits in other walks of life? Are we on a spiral of decision deferment to silicone processes? 'Computer says "No" '.
KC has hit on a good point. I witness people kibitz GM games, and they wonder how a GM missed a certain line, that perhaps their engine is seeing. Well, a lot of the times it's because the move that was played has a multitude of different ways it can win. Simple chess. Engines are greedy. They don't have any fear that they might not be noticing a tactic. So engines refrain from taking a piece in a 2-move combination if they find a 10-move combination where it will end up 1/100 a unit better.
I also think that's the difference between playing tactically and playing positionally. If you're playing positionally you're keeping a safe position where there are usually multiple good moves at any point of the game. Tactically you don't have that luxury so you have to play precise, accurate. A lot of the time there's only one good move.
The only engines I run are the freeware available on the net so I am somewhat engine ignorant. What is meant my the incremental depth? Does it meant that the computer is looking 17 moves ahead at depth 17? Or does it mean something else?
If this is the case shouldn't the computers be able to see the games in their entirety from move 1 and determine the best moves for both sides the whole way through? And, wouldn't this make the point system by which computers evaluate meaningless?
If a computer engine is at 'depth 17', my understanding of it is that the engine has analysed all the possible positions after 17 moves and has 'decided' what is the best play from both sides.
not true. If you do the math you will quickly understand that if this were the case, no computer would ever get to depth 17. The depth represents the longest line the computer has looked at, measured in plies. There will be many 3 ply deep positions that the engine has not looked at when it reaches depth 17, simply because it rejects these as inferior and thus does not waste time examining them.
exactly. The evaluation function is all that matters here. When the engine thinks that the line it is investigating is 'dead' it will turn its attention to other possibilities. The thing is that the exponential factor is so great in chess, that a huge amount of lines has to be cut as quickly as possible - sometimes leading to errors. This is visible to humans in endgames, where the human can establish a set of priorities and goals, whereas the engine still only got their evaluation.
I meant 3-move (6 ply) deep positions and not 3 ply. Still, this is probably an exaggeration. At depth 17 the engine will probably have examined all possible positions 8 or 9 plies deep.
Yes and no. They will never analyze "all" possibilities, they take reasonable guesses and take some shortcuts if you will - good shortcuts. If there is, for instance, a mate in X within the depth, you can be sure it will find it.
But otherwise they have a strong positional bias. Some kind of "horizon" problem still persists, if the gain is very long-term, or if at depth x+1 there is a very tactical move, they might miss it. They will though never blunder any move because of the horizon.
(ps: I was replying to jewbinson) and to finish my sentence: they will not blunder because of the horizon because they always make sure to finish all winning lines 'til all exchanges/checks/etc. are validated before scoring the position.
But not seeing an even more winning position is still very possible.
I think you're raising a great point here, although the presentation wasn't all that polished.
fateola 2 years ago
I found it a bit ironic that at depth 18 vs depth 17 black suddenly has to give up the rook on a8. And also Rybka doesn't give anything else to explain why. Black box =/
zaychenok 2 years ago
Thanks KC, very interesting video (as are your other ones) and your point about the mainline syndrome is interesting. My view on it is that yes, computers are better than humans at chess but if we start taking the computer's suggestion as the "correct" move, then our thinking will become misguided because we can't explain where this move comes from. Humans do have the ability to be innovative and create and refute hypotheses, but chess is a closed enough system to yield to brute-force search.
ytkonsta 2 years ago
Yes, you can choose line 4, but if your opponent plays the "best" moves following your moves, you will be in that much less of a position.
My comments assume that the opponent plays the "best" moves. Otherwise, line 5 could be beat a typical IM, etc.
bound4h 2 years ago
i think its cheesy that you backed off engine analysis because of a GM's comments. An engine's algorithm finds the quickest a most effective way to 1) gain piece advantage, and 2) mate the opponent. To assume that you can take the second line as opposed to the main line and not be a Mainline Servant and still be in a better position is silly. The mainline is just that, the main line of moves that will accomplish the above to goals the quickest. running out room to type...
bound4h 2 years ago
@bound4h
I think the point here is that eng analysis is more suited to analyzing games between engines. Engines miss the human aspect of analysis, and only best moves found by humans take this account into consideration.
So, the best move against an engine is not necessarily the best move against a human.
But I agree, the engine (these days) will be ultimately more correct even with subtle things like order of moves in a sequence. Engines are stronger ultimately than humans.
QuakePhil 2 years ago
you've heard it here first folks!
intindse 2 years ago
like a little noel
trixapete 2 years ago
Engines have no style, we still have that for a while.
StinkySQL 2 years ago
different engines got different evaluation functions and thus different styles.
Ruxistico 2 years ago
I'm really happy you tied zen and the art of motorcycle maintenance into this discussion! Engine analysis is a really thought-provoking application of Pirsig's ideas about subject-object interaction and the emergence of 'quality'.
Any chance you could post the video of that Short game as a reply? (btw, it's pretty cool that GMs are checking out the videos you do of their games, props man).
mindgrapes 2 years ago
I also believe that you're kind of overstating the obvious here KC, no offense. Anyone and everyone not only playing chess, but even further, using the engine to analyze games is certainly aware of the fact that what would normally be considered even brilliant by human standards, is most often brushed away by the likes of Rybka and others in favor of the 1,855,907th Var. If the sole point is to say, make sure you balance your play and realize that your opponents are only human too, then kudos.
JohnnySkillish 2 years ago
There is actually a deeper aspect about the lack of "internal quality" focus, which I will try and expand on in another video part to this. Without such a focus, it is as if we are just accepting the results of engines without question.
kingscrusher 2 years ago
KC,where can I buy Rybka and learn how to make chess vids?
Can you consider becoming sort of People's Enlightenment Guru and spread the light of knowledge in philosophy,psycology,natural sciences,web administration,computers,chess on youtube?
Can you promote chessboxing as well?
gologram 2 years ago
Thanks for the reply KC. I agree that it;s very difficult NOT to just accept a Rybka-like response, after all, it IS A COMPUTER, RIGHT?! Tongue-in-cheek. I'll have to watch your second installment here and respond...
JohnnySkillish 2 years ago
@JohnnySkillish Yes but it is like accepting an artefact of "external quality" without having access to its "internal quality". You don't understand that artefact -it is just a black box. So despite being technically great, it is also unevolvable, unexplainable, etc.
kingscrusher 2 years ago
Great analogy. I did watch the second installment, which was fascinating once again. I would like to read your response to it my post on that page if you are so inclined. And anyone else's too for that matter. Thanks again! JS
JohnnySkillish 2 years ago
So then, do the engines just go for the quickest mate possible as opposed to taking material off the board?
bpearsall0 2 years ago
So then, do the engines just go for the quickest mate possible as opposed to taking material off the board?
bpearsall0 2 years ago
I am always attempting to be cognoscente of the fact that when I use Rybka to analyze a game I've played, that BOTH sides are being analyzed and that the program is always going to be looking for equalization, which is why true sac's are never made and also why "simple" and "obvious" moves are often simply dispensed of by Rybka. Therefore, I insert other "human" moves based on my own thoughts in a position and see where it goes. Surprisingly, sometimes it actually ends up better in the end.
JohnnySkillish 2 years ago
I think a lot of people noticed your mistake at the end of that game's commentary. I know I was thinking the same thing after watching the position. All the moves are made to get to a position where the bishop can be captured. I suppose the lesson is that there needs to be preparation when commenting on games to avoid running into claiming a computer zwischenzug as a legitimate brilliant or winning move.
I'm looking forward to future commentary, especially if you stick by what you've said here
ShinobiAC 2 years ago
Very interesting discussion! I hope that it will still be alive when I return (since I am too tired to participate as of now). I have done extensive work on the subject (as part of a larger study of AI in general) and I am really glad that you brought it up!
Ruxistico 2 years ago
Very interesting dialogue there Tryfon, it reflects some things that i've said to friends before. I've always had the opinion that chess is about truth. Both players will see the position differently, have different ideas about what's going to happen next, and what's possible tactically, it's the person that sees the truth of the position better that is going to win.
The problem with engines, and society in general, is when people stop questioning the answers they're given. ;)
l3xrun 2 years ago
as you mentioned, zen and the art of motorcycle maintenance is a good book :)
HomelessTwit 2 years ago
Very interesting indeed. I wonder if 'Mainline Syndrome' exits in other walks of life? Are we on a spiral of decision deferment to silicone processes? 'Computer says "No" '.
ImposingSumo 2 years ago
KC has hit on a good point. I witness people kibitz GM games, and they wonder how a GM missed a certain line, that perhaps their engine is seeing. Well, a lot of the times it's because the move that was played has a multitude of different ways it can win. Simple chess. Engines are greedy. They don't have any fear that they might not be noticing a tactic. So engines refrain from taking a piece in a 2-move combination if they find a 10-move combination where it will end up 1/100 a unit better.
mschonert7 2 years ago 2
I also think that's the difference between playing tactically and playing positionally. If you're playing positionally you're keeping a safe position where there are usually multiple good moves at any point of the game. Tactically you don't have that luxury so you have to play precise, accurate. A lot of the time there's only one good move.
mschonert7 2 years ago
the genius of man is beyond the genius of computer!
buji81 2 years ago 2
The only engines I run are the freeware available on the net so I am somewhat engine ignorant. What is meant my the incremental depth? Does it meant that the computer is looking 17 moves ahead at depth 17? Or does it mean something else?
If this is the case shouldn't the computers be able to see the games in their entirety from move 1 and determine the best moves for both sides the whole way through? And, wouldn't this make the point system by which computers evaluate meaningless?
DriftingAimlessly 2 years ago
If a computer engine is at 'depth 17', my understanding of it is that the engine has analysed all the possible positions after 17 moves and has 'decided' what is the best play from both sides.
jewbinson 2 years ago
not true. If you do the math you will quickly understand that if this were the case, no computer would ever get to depth 17. The depth represents the longest line the computer has looked at, measured in plies. There will be many 3 ply deep positions that the engine has not looked at when it reaches depth 17, simply because it rejects these as inferior and thus does not waste time examining them.
Ruxistico 2 years ago
is this why sometimes it will miss, for example, a queen sac which is 7 or 8 , moves deep?
jewbinson 2 years ago
exactly. The evaluation function is all that matters here. When the engine thinks that the line it is investigating is 'dead' it will turn its attention to other possibilities. The thing is that the exponential factor is so great in chess, that a huge amount of lines has to be cut as quickly as possible - sometimes leading to errors. This is visible to humans in endgames, where the human can establish a set of priorities and goals, whereas the engine still only got their evaluation.
Ruxistico 2 years ago
I meant 3-move (6 ply) deep positions and not 3 ply. Still, this is probably an exaggeration. At depth 17 the engine will probably have examined all possible positions 8 or 9 plies deep.
Ruxistico 2 years ago
Yes and no. They will never analyze "all" possibilities, they take reasonable guesses and take some shortcuts if you will - good shortcuts. If there is, for instance, a mate in X within the depth, you can be sure it will find it.
But otherwise they have a strong positional bias. Some kind of "horizon" problem still persists, if the gain is very long-term, or if at depth x+1 there is a very tactical move, they might miss it. They will though never blunder any move because of the horizon.
alxborgia 2 years ago
(ps: I was replying to jewbinson) and to finish my sentence: they will not blunder because of the horizon because they always make sure to finish all winning lines 'til all exchanges/checks/etc. are validated before scoring the position.
But not seeing an even more winning position is still very possible.
alxborgia 2 years ago
Very interesting. Could you provide a link to the correspondence between you and Nigel Short, so we can read it in its entirety?
111Socrates777 2 years ago