I agree with the fundamental premise but think it maybe misses the point. A set of principles cannot tell us whether a4 is better than b4 in that position, but they do help remind us of what elements we should consider when comparing the moves from a strategical perspective. Eg there are outposts, backward pawns, etc, and these are probably factors that we should think about when choosing what move to make. Eg, is the outpost usable (and useful) by us or our opponent? Is the pawn vulnerable or can we swap it off anyway? Principles can't answer these concrete questions but they can help us in finding the questions to pose.
Of course an immediate tactical refutation of either move trumps strategic considerations anyway.
@DailyInsanity said in #8:
> However, I struggle to see how principles could help in variants like crazyhouse, according to the same line of argumentation as in the post. I'm unsure about other games since I don't know much about them.
That's a very good point. It's true that tactical vision goes a long way in crazyhouse, so general principles tend to be about looking out for weaknesses on particular squares, not bringing the queen out early, watching out for pawn forks, etc. On the other hand, shogi proverbs are widely acknowledged to be useful since the game is more strategic and less based on short-term tactics.
Reduction needs recombination (or superposition). And that is not the mere addition of the reduced isolated parts.
SF classicla assumed linear combination for a long time, as the formula for hindsight born then used forward in their engines, human chess theory elements (why many people might be mourning that it is now being recycled into a more information preserving but more obscure (in one way but not others) NN taking charge of the superpositoin problem.
And it seems, they even have their own Elo worth hypotheses going on. I say, do not mourn, they are being sifted through to their efficient cooperation as input space dimensions. There might still be the mysterious magic bullet piece square table transformation clone, the common DNA of all the engine not of the clade of A0/LCO full chess input space using NN with all their teeth still in use (costly).
But that is another charade, (I just had the thought that it might be an emergent truth, despite all the "random" (not formalized for generalization problem global optimization) training scheme across SF versions, until A0/LC0 knocked that in some other direction.
I wondered how a full chess problem formalism like the A0/LC0 architecture (thinking more of the mathematical model I digested from reading things, not code), would transform certain things to make its last head reward dense decision maker, to see a better/clearer evaluation method (not linearly separable of sorts, although this is regression at that point, it might be about contours more than separation).
Could the ad-hoc magically found (or perhaps individual dev intuition toying with engine with some intuition going on while manually exploring parameters (I have no clue, I wish it had been shared, but all I know is how contagious it is in all sort of engines of the type of SF.
One thing one can say, and this is off-topic sorry (but tangent), is that it must make all the engines in that group playing better against each other. But, if that has sustained invaders like A0 and LC0 in the pool, then it might hold generalizatoin potential over the unknown union of all the position visited by all the engines playing with each other in history. or still stil with each other (not sure, the role of history, but SF is keeping regressions pools).
now back closer. That was improvized last paragraph. To share my point of view that also is applicable in human chess, I think. as model on our now, part of discourse intution shenanigans, that an analogous piece square table transfomatoin would have to be for our conscious minds not to start freezing, and us drool all over the board.
how the non complete human genious can still find things before it even knows how it could find it methodically. Yet it must have. Still, not communicable until tested that kind of thinking of mine. (hoping not just mine).
I'm still convinced that principles, and knowing why, are useful. The cherry-picked example I'll use is on how to win a king+rook endgame vs a lone king. For example, if white has a king on e4, rook on d5, and black a king on b6 with white to move. As a kid I was taught to move the rook to h5 here, trying to move the rook to h6 whenever the white king takes opposition, while chasing it to force it to take opposition. Using this principle, it's forced mate in 21 moves. I've since learned faster ways and different principle that would have me play different moves here (moving the king closer). However, that principle would've been riskier to teach me as a kid because I could more easily hang the rook with a wrong move. So, I can agree with your two premises:
- Different principles give you different moves to play: the principle you follow influences the decision making promise
- In this particular position there are different principles, there is no objective criterion for a "best principle" (one leads to faster mate, the other is harder to blunder with)
It is also clear that your conclusion 2 doesn't hold: we can use either principle and it doesn't matter which, as long as we stick to it (if we'd mix principles, I could be repeating moves). Conclusion 3 and 4 also do not hold, they are in direct conflict with the first premise.
I'll admit: this position was cherry-picked. Most positions are harder to explain, and don't have an objective evaluation (just a computer's best guess). However, it shows that your conclusions do not follow from your premises, at least for the principles I cherry-picked. My current belief is that principles are useful still, and at all levels even, it just becomes harder to communicate as at higher levels, the group who can understand certain principles shrinks while the set of principles used is growing. But I envision that even in preparation of world championships, ideas in certain positions are formulated as principles: I imagine they'd be saying something like "If black ever plays Nf6, you want to look if you can transpose into one of the open lines, so that's why you keep the pawn tension." which would then be a principle that only really applies for the duration of five or six moves, and only because in the preparation for this particular game the player looked at certain "open lines".
Euclid is reputed to have said, "There is no royal road to geometry." We might also say there is no royal road or easy shortcut to chess mastery. Since philosophy was mentioned, especially syllogisms and pure reasoning, we need to consider some insights from Francis Bacon. Let us start with two basic definitions.
1. Syllogism - An instance of a form of reasoning in which a conclusion is drawn from two given or assumed propositions (premises). A common or middle term is present in the two premises but not in the conclusion, which may be valid or invalid.
Example: "All men are mortal. Socrates is a man. Therefore, Socrates is mortal."
We will assume this was said when Socrates was alive. It as an example of a correct or valid syllogistic deduction. The two premises are correct, therefore the deduction is correct. This illustrates both the strength and weakness of such deduction. If one of the premises is false then the deduction will be false. If both of the premises are false, the deduction is likely to be false but it might be fortuitously true because of the double negative logic. I haven't looked into this,
2. Pure Reasoning - Reason based on a priori principles (premises) and providing a unifying ground for perception and analysis of the world or a part of it.
What normally gets opposed to pure reason? I would say that the opposed principle is empiricism or "knowledge from experience".
Francis Bacon was perhaps the first serious Western critic of syllogistic reasoning in application to natural philosophy questions. Natural philosophy or philosophy of nature (from Latin philosophia naturalis) is the philosophical study of physics (and of chemistry and biology), that is, of nature or the material universe. Bacon came to be regarded, in the West, as the father of modern science and of the scientific (empirical) method. His aphorisms on this matter are concise and profoundly clear. I can give some connected examples of this. Then I can lead on to a discussion of the collision of formal systems and real systems and how the brain might deal with formal systems. For chess is a formal system, like geometry for example, and the human brain (and its processes and its situation in a nervous system in a human body) is a real system. I can also deal, briefly, with cultural attempts, like the Soviet System at its height, at creating grand master chess players. Their system was highly successful, at the time and at its height. That is an empirical result. Was it due to the principles of play working or the principles of selecting raw talent from a population working? Or a mix? All these questions must be considered.
Finally, in evolutionary terms, humans and their precursor species evolved to deal with the physical world (the physis) first and only later started evolving to deal with the cultural world (the nomos) and its systems of formal signs. The evolution of the "plastic brain" has a lot to do with both of these issues. Such observation has a lot of implications. But I will not expand more on this unless I pique interest.
Footnote: I am a really bad chess player so should I be commenting at all?
I am not sure who of the blog or the discussion post above, is saying "it does not matter which of 2 plausible principle based paths are taken, as long as we stick to it".
Well, if we are going the pragmatic way, and consider as part of the discourse the internal model in evolution of the improver/learner (any pace priority) for their own play given the board and some opponent in some set of opponent difficulties (I don't want to get lost in that vague notion here, it might itself be another charade to model, let's just go with the usual: rating, it is also vague but it means something conventionally, or within a pool, about average difficulty per position in some game event).
Then it the question of experience set dependent learning (where we consider explicitely the set of positions in the training set, and possibly as a time series, considering appropriate likely familiarity intensity over all of them, a function of position information, I would suggest), and we accept the possibility that familiarity might affect both evaluation in mindes eye including current position, then that obligatory experience set biased view of how to keep going, might be an argument to stick with it, as likely to be able to navigate better.
btw. that is also why I find the exageratoin of the 50 move rule, slippage into a shorter mate is better, to be conflicting with this idea.
And it would fit with machine learning models as well. In all of these discussion, there is still a hidden or floating variable, I claim, or intuiti, or hunch, and I could develop using machine learning inspired arguements, which is about the currently dismissed to black box intuition notion of similarity (and its more interesting and complete version, less often used, distance, so we know where similarity ends and distance start, or wehn a referential cluster center becomes close than the one we were measuring the similarity as only notion of distance).
In axiom1. Training set "diameter" and large set size? should we just count them. but then, every position that differs even by null moves or non causative change on odds would count as contributing to the learning load, if eventually we would have a theory of learning that could give some sense of the lifelong time commitment to get from somewhere to elsewhere.
Also in that line, of pragmatism and acknowledgement of some cognitive (hear learning) science, that sensory experience specifics, might affect the performance (including reading or hearing about chess theory I would add, as this is still not a closed discussion, it is). It might depend on the question of generalization. How does one measure it?
Again. We need some notion of how much has our possibly random repetition of a single chess task (game or mini-game down to even a mere static call), has been stagnant regarding the larger set (wilderness, as I seem to favor). Counting? No, it seems in our implicit mind we have been thinking space and distance. At least that is what I hear when someone is make similarity calls, and myself putting a tree of decision in minds eye (to conform to the tools), even the graph of that tree, in mind'eye takes some kind of 2 space (and the fake distance, that represent chess ply, is also kind of easily used with our distance notion, including my curiosity for the transversal directoin, which is not a legal move traversal idea, I am still using, perhaps or others are, that visual support, but in effect in the branching direction, it is just counting distance. And then how one chose to order the sibling branches might also be about counting, but this one would not have any ches interpretation (unless doing the silly thing that our tools force us, bracktrack to common ancestor, shift gear and then, ... where was I? no really where was I in the train of thought about the chess board? or the experimental design for my study progame (or theory), say, woodpecker (although I don't know exactly what that is). We need something that can be shared and not a secret of expertise. I decree.
Edit: did I say null move. Sorry, it keeps coming out wrong. I meant null-odds move. When it don't matter whose side has the turn, for example, which is usually only clear with our working memory is having the help of our spatial longer terms associative (well my word, but however our spatial intuition is implemented, it is not of the conscious traffic jam, or reduced slow bandwidth kind, it frees the working memory, if I were to step on someone's jurisdiction of science and gives some belief to such high-level model (for associative memory, I have neurobiological past knowledge, in support).
@dboing said in #20:
> Exactly.. So glad this crowd of thinkers can uncover things that could have been long ago. If we lifted our heads sooner.
That sure is an interesting way of putting it :) However, I want to clarify that I’m not asserting that this is true, but rather that it is plausible. In the post, I touch on this under the section titled "Chess Principles as a Temporary Tool." (On second thought, maybe I should have left it out since it raises an issue unrelated to the main argument of the post). As I say there, I believe the burden of proof lies with principle-supporters to demonstrate why this is true.
The potential reasons, such as those given by @MrMatt96 and myself, serve as starting points but remain largely assertions. At this stage, believing one side or the other seems to come down to a hunch rather than any well-defined line of reasoning. However, given the black box nature of evaluative intuition, I'd imagine it's quite hard to outline such a line of reasoning without some experimental/data-driven aspect.
For those that have this hunch (that chess principles are useful catalyzers of evaluative intuition), one aspect to consider is that we are still in the very early stages of understanding how chess intuitive evaluation develops. The idea that principles may catalyze the development of our intuition might hold because principles are among the few (only?) tools we currently have. In the future, with a more accurate understanding of intuitive evaluation and more advanced tools and technology to develop it, the role of principles in its development may seem quite minor (That is, assuming it does help to some extent in the first place).
I should also clarify that my objective isn't to suggest that we should completely disregard principles and never speak of them again (as some other commenters seem to imply). Rather, my aim is to highlight the over-emphasis we often place on principles, and specifically show how they cannot help us make decisions between a set of candidate moves.
Even if we were to assume that chess principles have absolutely no connection to practical chess, there is still value in having them. As I discussed back in part 1 of this series, principles can be useful from an academic and theoretical perspective. They are interesting to think about and contemplate, just like understanding and thinking about Mbappe's technique is also interesting (the analogy I used in part 1).
@jdannan said in #21:
> I agree with the fundamental premise but think it maybe misses the point. A set of principles cannot tell us whether a4 is better than b4 in that position, but they do help remind us of what elements we should consider when comparing the moves from a strategical perspective.
I think that raises an excellent point: while principles may not directly dictate exactly which move we should select, they can highlight specific aspects of the position that influence our decision-making process in other ways. I do believe this is true, but only in a very specific context.
In part 3 of the series, I outlined that the decision-making process in unknown positions consists of two components: seeing and evaluating. With this current post, I aimed to show that principles cannot help with the evaluation component of the decision-making process. However, I do believe that sometimes, certain principles can help us discover previously unconsidered moves (I call these principles 'vision tools,' and I plan to discuss them in future posts).
>Eg there are outposts, backward pawns, etc, and these are probably factors that we should think about when choosing what move to make. Eg, is the outpost usable (and useful) by us or our opponent? Is the pawn vulnerable or can we swap it off anyway? Principles can't answer these concrete questions but they can help us in finding the questions to pose.
I can understand that line of thinking, but my issue is that when we further investigate how such observations influence the 'evaluating' half of the decision-making process, it's challenging to pinpoint exactly how they help. I think anyone supporting this view would need to be more precise in outlining exactly how considering such factors can lead to a player changing their initial move preference, given the argument I presented in the post.
fair enough.. I think one does not change perspective during a discusson. or even many. . And there is some value in sticking to it, until we feel like all our hunches are made communicable, with the help of common rules of verbal argumentation. I am sorry by the way, for some styling I use to keep the enthousiasm of verbal communicatnoi going, in spite of its flaws, when talking about complex things, I also have frustrations from compromising and conforming to things that felt a bit contorted by traditions or just me not being part of the group.. And I get upsurges of court jesterism. I might be making some statments about axious improvers, that might not be applicable to all ambitious improvers. And now, I find I need to defend the possibly that it migh be part of the chess improver culture, the exageration I used. So if not the reader case, I am sorry, and if near it, maybe consider breathing and enjoying the exploraton of the chess board mysteries too.. (for your comment about nobody being curious about chess that way, which was winking, and so am I , or Am I?) (kidding). I am far from done with forward reading. I give myself many days. to do so, given my resources.
TBH or thinking back. I also do not know for sure. But I sense that we can't say either way. And since I am still a social animal, and I like my humankind where it wants in good faith to share their love of chess, and their theories.
In doubt, I let my instinct some lea way, and it means discussion, debate where the common search for better understanding dominates the self-image engine we are chained to for life (and seems to fly off handle easily, in social competitions pairwise activities with some visibility, I find chess theory a bit soothing that way). Half kidding.
I salute your style of blog and debating, by the way. Not at all within my current pre-conceptions, I may have given a hint of above. But yes, thank you for replying, I went a bit overboard, showing my bias toward some iconoclastic hypotheses, but calming down, indeed, it should be held as a hypothesis. Just that we have not much else going in on as shareable tools, and I am not getting younger, even if I was in the market for the reduced target of theory of improvers.
But subjectively, as possibly outlier learner, I have manage a healthy skeptical self-centered wheel of science in my correspondance chess, when I can, and have the presence of mind or energy on enough days during the long lasting, and imprinting, games (including those that don't make it to the lichess record, on positions that strike my hesitations, i.e. where I might be on the edge of my familiriaty groove, I am impulsively attracted to those...).
And really, applying your reasonning method, and trying to scrape some fog off the existing chess theory resources, has given me a somewhat rational impression of steadily augmenting understanding, and I might call that of the board logic, superposed to the turn by turn tactical views (in spite of the dichotomy being a scaffold, for our small brains needing chunks in all directions). ok. I stop.. better read . I even have not read forward the posts I reacted to..