Stockfish NNUE style

[MikeB]

It would not work well - I agree . Seems like a useless exercise anyway - since the point of using NN is to ge to the a more accurate of the truth and what they want, is an engine that plays a certain type of fiction with style. Good luck with that... - they can let me know how that out works for them ...

So, you would seem to think that there is truly no 'style' in chess?
...

Did i say that? No I did not , you are reading the thread out of context and misinterpreted what i meant and perhaps i did not say it well enough. Anyway, I am moving on.

Move on.
You were not clear so I simply asked.

The phrase "...what they want, is an engine that plays a certain type of fiction with style" Very opague....the a phrase 'type of fiction with style', is probably a phrase that has never been uttered in the history of mankind....that I am aware of.
Anyway, by saying "the point of using nn is to ge(t?) to a more accurate (more accurate what?) of the truth... was poorly written and just sounded like maybe you were trying to indicate 'style' was not possible, only 'the truth'...whatever that is.
Sorry....hope you can see now why as I asked that.

I agree - I could have wrote it better - generally speaking, chess is a game of truth - you might one best move or several moves that are best - but every position has a best move or a a few best moves. NNUE net development is an attempt to get the truth for each position, there is no style in getting at the truth. Any diversion from getting at the truth , is no longer truth , it is fiction, meaning it is false, it is not the best move. Getting an engine to play with style -which may be possible - certainly humans play with style as they try to play for the best move - humans will tend to play with what has worked for them before - but it may not always be best move.

I deal with making engines play weaker all the time, it is hard to make them play weaker and have them playing at a lower level and have it not look artificial. Hence my comment , "good luck with that". I wish them well.

[Albert Silver]

So long as we define perfect chess as chess moves that always achieve the best possible result from a given position (i.e. if it is winning then perfect chess would not fail to win it), there is a ton of room for style while being perfect.

Exploitative play will vary according to the opponent, so even play that provides the best chances for a non-perfect player to lose will vary. The move that best trips Kasparov may very well not be the best move to trip Karpov after all.

So assuming that the starting position is a draw objectively, all the moves and plays that do not sacrifice this can easily be chaotic madness, or quiet maneuvering.

[syzygy]

So long as we define perfect chess as chess moves that always achieve the best possible result from a given position (i.e. if it is winning then perfect chess would not fail to win it), there is a ton of room for style while being perfect.

Exploitative play will vary according to the opponent, so even play that provides the best chances for a non-perfect player to lose will vary. The move that best trips Kasparov may very well not be the best move to trip Karpov after all.

So assuming that the starting position is a draw objectively, all the moves and plays that do not sacrifice this can easily be chaotic madness, or quiet maneuvering.

I fully agree!

The "more perfect" engines get, the more important it becomes for them to "understand" the weaknesses of the opponent. Playing chess is not about playing perfect moves but about luring the opponent into making a mistake (saying this somehow offends many people, I have noticed in the past).

[Cornfed]

I agree - I could have wrote it better - generally speaking, chess is a game of truth - you might one best move or several moves that are best - but every position has a best move or a a few best moves. NNUE net development is an attempt to get the truth for each position, there is no style in getting at the truth. Any diversion from getting at the truth , is no longer truth , it is fiction, meaning it is false, it is not the best move. Getting an engine to play with style -which may be possible - certainly humans play with style as they try to play for the best move - humans will tend to play with what has worked for them before - but it may not always be best move.

I deal with making engines play weaker all the time, it is hard to make them play weaker and have them playing at a lower level and have it not look artificial. Hence my comment , "good luck with that". I wish them well.

Thanks for the clarification.

It strikes me that 'style' in human chess and 'engine chess' are rather different animals. In human chess, we see more of a 'game of mistakes' (big and small) as we humans are not 'perfect'. Therefore Carlsen may play a "quite opening" with White (which an engine's eval might not find 'appropriate') and look to exploit the human side of his opponents with his natural ability to make something out of (often) next to nothing. A 'bookish' player, might stick to the relatively narrow streets he is comfortable traversing while someone who excels in tactics might branch off a line of play towards positions where his opponent needs to be careful to not fall into a worse position which might snowball to even worse positions.

The natural talent of a Carlsen may give him more chances ultimately because he takes play off into a direction where he excels....and avoids so many hours of preparation by many opponents who have had seconds or engines work out all the critical moves in the lines they play. That he can be a very good attacker and isn't afraid to switch 'style' when given the option certainly does not hurt.

I can see where it would be hard to program 'style' into an engine. We have all seen Karpov patiently nourish his game (in a totally different way than a Kasparov or Tal might) to a certain point and then cap his play off with a 'change of personality'...as if 'style' were but a point on a sliding scale or drift chart where he continues as he sees fit or is simply most comfortable with - sharply or less so. An engine is going to pick it's eval of a +.70 position over a +.62 or .57 position regardless of if the .70 is less likely to offer its opponent more opportunities to 'go wrong' to some degree over the next series of moves. This is perhaps why Human chess is seen more as a 'game' and engine chess more a 'mathematical exercise'?

That's what engines (well, the programming) seem to lack - true personality or choice in 'style' as the programmers tend to look for this chimera (?) known as 'truth'. It is why Komodo MTCS and now ShashChess and to a degree the NN's intrigue me. Although I've read the Shashin book (it was a gift I forced myself to read) and and have never really seen a good exposition (just generalities) of how it's play or choice of moves correlate with the book.

[chrisw]

So long as we define perfect chess as chess moves that always achieve the best possible result from a given position (i.e. if it is winning then perfect chess would not fail to win it), there is a ton of room for style while being perfect.

Exploitative play will vary according to the opponent, so even play that provides the best chances for a non-perfect player to lose will vary. The move that best trips Kasparov may very well not be the best move to trip Karpov after all.

So assuming that the starting position is a draw objectively, all the moves and plays that do not sacrifice this can easily be chaotic madness, or quiet maneuvering.

I fully agree!

The "more perfect" engines get, the more important it becomes for them to "understand" the weaknesses of the opponent. Playing chess is not about playing perfect moves but about luring the opponent into making a mistake (saying this somehow offends many people, I have noticed in the past).

MCTS is by its averaging nature a luring algorithm. It seeks regions of the tree where it tends to find many “winning” situations for itself and conversely many “opportunities to go wrong” for the opponent. AB on the other hand is fine with picking out one fine and select path just so long as it likes the position at the end of the line. MCTS seeks to take you into a minefield, AB seeks to come out the other side of the minefield in one piece (or rest on an apparently safe stepping stone if it cant find apparent quiescent safety).
Luring the opponent into making a mistake OTB (setting traps) is either a very high level playing skill, or simply reflects big disparity between players. It’s pretty dangerous to assume you saw the trap while the opponent doesn’t, assuming your opponent is equally strong player. It probably also assumes “tactics” and in most high level chess, tactics is what you finish off positional advantage with, imho.

[mwyoung]

So long as we define perfect chess as chess moves that always achieve the best possible result from a given position (i.e. if it is winning then perfect chess would not fail to win it), there is a ton of room for style while being perfect.

Exploitative play will vary according to the opponent, so even play that provides the best chances for a non-perfect player to lose will vary. The move that best trips Kasparov may very well not be the best move to trip Karpov after all.

So assuming that the starting position is a draw objectively, all the moves and plays that do not sacrifice this can easily be chaotic madness, or quiet maneuvering.

I fully agree!

The "more perfect" engines get, the more important it becomes for them to "understand" the weaknesses of the opponent. Playing chess is not about playing perfect moves but about luring the opponent into making a mistake (saying this somehow offends many people, I have noticed in the past).

MCTS is by its averaging nature a luring algorithm. It seeks regions of the tree where it tends to find many “winning” situations for itself and conversely many “opportunities to go wrong” for the opponent. AB on the other hand is fine with picking out one fine and select path just so long as it likes the position at the end of the line. MCTS seeks to take you into a minefield, AB seeks to come out the other side of the minefield in one piece (or rest on an apparently safe stepping stone if it cant find apparent quiescent safety).
Luring the opponent into making a mistake OTB (setting traps) is either a very high level playing skill, or simply reflects big disparity between players. It’s pretty dangerous to assume you saw the trap while the opponent doesn’t, assuming your opponent is equally strong player. It probably also assumes “tactics” and in most high level chess, tactics is what you finish off positional advantage with, imho.

So logically chess is all tactics, and positional advantage is just knowing tactics will win in the end. Without the ability to see the tactics to the end.

[chrisw]

So long as we define perfect chess as chess moves that always achieve the best possible result from a given position (i.e. if it is winning then perfect chess would not fail to win it), there is a ton of room for style while being perfect.

Exploitative play will vary according to the opponent, so even play that provides the best chances for a non-perfect player to lose will vary. The move that best trips Kasparov may very well not be the best move to trip Karpov after all.

So assuming that the starting position is a draw objectively, all the moves and plays that do not sacrifice this can easily be chaotic madness, or quiet maneuvering.

I fully agree!

The "more perfect" engines get, the more important it becomes for them to "understand" the weaknesses of the opponent. Playing chess is not about playing perfect moves but about luring the opponent into making a mistake (saying this somehow offends many people, I have noticed in the past).

MCTS is by its averaging nature a luring algorithm. It seeks regions of the tree where it tends to find many “winning” situations for itself and conversely many “opportunities to go wrong” for the opponent. AB on the other hand is fine with picking out one fine and select path just so long as it likes the position at the end of the line. MCTS seeks to take you into a minefield, AB seeks to come out the other side of the minefield in one piece (or rest on an apparently safe stepping stone if it cant find apparent quiescent safety).
Luring the opponent into making a mistake OTB (setting traps) is either a very high level playing skill, or simply reflects big disparity between players. It’s pretty dangerous to assume you saw the trap while the opponent doesn’t, assuming your opponent is equally strong player. It probably also assumes “tactics” and in most high level chess, tactics is what you finish off positional advantage with, imho.

So logically chess is all tactics, and positional advantage is just knowing tactics will win in the end. Without the ability to see the tactics to the end.

No.

[chrisw]

It would not work well - I agree . Seems like a useless exercise anyway - since the point of using NN is to ge to the a more accurate of the truth and what they want, is an engine that pla OO ys a certain type of fiction with style. Good luck with that... - they can let me know how that out works for them ...

So, you would seem to think that there is truly no 'style' in chess?
...

Did i say that? No I did not , you are reading the thread out of context and misinterpreted what i meant and perhaps i did not say it well enough. Anyway, I am moving on.

Move on.
You were not clear so I simply asked.

The phrase "...what they want, is an engine that plays a certain type of fiction with style" Very opague....the a phrase 'type of fiction with style', is probably a phrase that has never been uttered in the history of mankind....that I am aware of.
Anyway, by saying "the point of using nn is to ge(t?) to a more accurate (more accurate what?) of the truth... was poorly written and just sounded like maybe you were trying to indicate 'style' was not possible, only 'the truth'...whatever that is.
Sorry....hope you can see now why as I asked that.

I agree - I could have wrote it better - generally speaking, chess is a game of truth

so say most of the fundamental religionists who gravitate to computer chess. Actually, it’s no more about “truth”, whatever that means in this context, than any other other zero-sum two player game, tic-tac-toe, othello, shogun (edit: iPhone spelchek) or any other.

- you might one best move or several moves that are best - but every position has a best move or a a few best moves. NNUE net development is an attempt to get the truth for each position, there is no style in getting at the truth. Any diversion from getting at the truth , is no longer truth , it is fiction, meaning it is false, it is not the best move.

parallels with only-one-god religious belief systems jumping off the page here.

It always intrigued me, why so many fundamentalist Christians as users, also programmers, in computer chess? In all my other existences I come across absolutely zero, maybe a Jehovah’s Witness every couple of years knocking on the door. And why computer chess? I always figured something to do with too complex to understand, whilst appearing to have some kind of unachievable answer, or “truth” as you seem to call it. Anyway, whatever, and so on.

Getting an engine to play with style -which may be possible - certainly humans play with style as they try to play for the best move - humans will tend to play with what has worked for them before

curious generalisation. Do they? They may have desired openings, but beyond that ..,

- but it may not always be best move.

that thing again.

I deal with making engines play weaker all the time, it is hard to make them play weaker and have them playing at a lower level and have it not look artificial. Hence my comment , "good luck with that". I wish them well.

[Cornfed]

So long as we define perfect chess as chess moves that always achieve the best possible result from a given position (i.e. if it is winning then perfect chess would not fail to win it), there is a ton of room for style while being perfect.

Exploitative play will vary according to the opponent, so even play that provides the best chances for a non-perfect player to lose will vary. The move that best trips Kasparov may very well not be the best move to trip Karpov after all.

So assuming that the starting position is a draw objectively, all the moves and plays that do not sacrifice this can easily be chaotic madness, or quiet maneuvering.

I fully agree!

The "more perfect" engines get, the more important it becomes for them to "understand" the weaknesses of the opponent. Playing chess is not about playing perfect moves but about luring the opponent into making a mistake (saying this somehow offends many people, I have noticed in the past).

MCTS is by its averaging nature a luring algorithm. It seeks regions of the tree where it tends to find many “winning” situations for itself and conversely many “opportunities to go wrong” for the opponent. AB on the other hand is fine with picking out one fine and select path just so long as it likes the position at the end of the line. MCTS seeks to take you into a minefield, AB seeks to come out the other side of the minefield in one piece (or rest on an apparently safe stepping stone if it cant find apparent quiescent safety).
Luring the opponent into making a mistake OTB (setting traps) is either a very high level playing skill, or simply reflects big disparity between players. It’s pretty dangerous to assume you saw the trap while the opponent doesn’t, assuming your opponent is equally strong player. It probably also assumes “tactics” and in most high level chess, tactics is what you finish off positional advantage with, imho.

Agree...100% I guess!
But I would maybe argue with "Luring the opponent into making a mistake OTB (setting traps) is either a very high level playing skill, or simply reflects big disparity between players. It’s pretty dangerous to assume you saw the trap while the opponent doesn’t, assuming your opponent is equally strong player."

So logically chess is all tactics, and positional advantage is just knowing tactics will win in the end. Without the ability to see the tactics to the end.

Pretty much agree with that.

I would maybe nitpick with "Luring the opponent into making a mistake OTB (setting traps) is either a very high level playing skill, or simply reflects big disparity between players. It’s pretty dangerous to assume you saw the trap while the opponent doesn’t, assuming your opponent is equally strong player."

I suppose it depends on what once considers a "very high level of playing skill".

In my case I am only interested in using an engine to help maximize my chances in OTB or online play. Therefor, I would use MCTS for it's better multi-PV abilities. For instance, I play a sharp Scandinavian rather often, especially in quick play. If I am looking at a line that on move 12 offers me a +.1 advantage but there are 3 other lines that offer less or even a -.33 disadvantage, I might well opt for the 'worse' recommendations. This because I can more easily spot where those 'worse' lines might require my opponent to go thru a minefield....when the 1st move choice might not lead to much at all.

Human chess is like that and 'for sport' I love watching high level tourney and people on chessbomb who are not very good look at the lines shout 'blunder' or 'bad move' when they see red...only to see the person who might be down by .75 end up outplaying their opponent. Chess as played by humans is not accounting (or bean counting) it's about results.

[mwyoung]

So long as we define perfect chess as chess moves that always achieve the best possible result from a given position (i.e. if it is winning then perfect chess would not fail to win it), there is a ton of room for style while being perfect.

Exploitative play will vary according to the opponent, so even play that provides the best chances for a non-perfect player to lose will vary. The move that best trips Kasparov may very well not be the best move to trip Karpov after all.

So assuming that the starting position is a draw objectively, all the moves and plays that do not sacrifice this can easily be chaotic madness, or quiet maneuvering.

I fully agree!

The "more perfect" engines get, the more important it becomes for them to "understand" the weaknesses of the opponent. Playing chess is not about playing perfect moves but about luring the opponent into making a mistake (saying this somehow offends many people, I have noticed in the past).

MCTS is by its averaging nature a luring algorithm. It seeks regions of the tree where it tends to find many “winning” situations for itself and conversely many “opportunities to go wrong” for the opponent. AB on the other hand is fine with picking out one fine and select path just so long as it likes the position at the end of the line. MCTS seeks to take you into a minefield, AB seeks to come out the other side of the minefield in one piece (or rest on an apparently safe stepping stone if it cant find apparent quiescent safety).
Luring the opponent into making a mistake OTB (setting traps) is either a very high level playing skill, or simply reflects big disparity between players. It’s pretty dangerous to assume you saw the trap while the opponent doesn’t, assuming your opponent is equally strong player. It probably also assumes “tactics” and in most high level chess, tactics is what you finish off positional advantage with, imho.

So logically chess is all tactics, and positional advantage is just knowing tactics will win in the end. Without the ability to see the tactics to the end.

No.

Then lets get the the heart of the question. What is tactics vs positional play? And what is the difference between the two, other then the distance to calculate the line fully, and to use experience.

I am still open to the question!

When does tactics end, and positional play began?

Educate me.