so we dissect certain things but not others. how about dissecting the word principle?
did all the posts and the author, talk about the same things. has there been dialogs that passed through each other, or loops? I thought I read some of that. An objective board feature in itself is not a princiople.
it is even a seeable. I think the "from scratch" should be even handed. Care about all the buzz words that have been overused without cirtical thinking from the obedient students receiving the recipes for glory, generation after generation, and many books later. At least that which is being critiqued should get a fair equilvalent effort as that which is presented as a replacement. With the possiibility that one is completely enough total replacement. (the MFTL pamphlet).
I was enthoused until I saw that the same care was not deployed. it might be my exhausted reading which I am kind of giving up on.. I surrender. to my limitiations.
I can only discuss right now. and it is too late.. for previous posts. also exhausted for that. but if the above sub topic does resonate. I am willing to keep reading here. eventually another way, I might figure out what I could not read forward. sorry. I was not lazy like that. many years ago. and I should have known by now, what I can and can't chew.
@sjcjoosten said in #36:
> Here there's perhaps a misunderstanding: I argued that for the two principles and the particular position, the premises do hold, just not all of the conclusions.
Oh I see. I think I was under the assumption that you thought the conclusions didn't hold because one or more of the premises was false. In that case, I'm not sure why you believe the conclusions are false.
Reading over your first message again, perhaps you say that conclusion 2 is false because in "We cannot use P1, P2, P3 etc... when determining which of m1, m2, m3 etc... should be preferred." you are taking 'preferred' to mean the preferred move as in the best move? I was meaning 'preferred' in the sense of the move recommended by a principle, independent of the objective status of that move. If a principle recommends move A, then that is the preferred move if we accept that principle. But since we have multiple principles, that are recommending different moves, there is no way to arrive at the 'ultimate' preferred move. In this sense it doesn't matter that in your particular example, both moves happen to lead to forced mate, only that principles still cannot dictate which path we choose.
This conclusion then leads to conclusion 3, which combines with the premise 1, a conditional statement, to arrive at conclusion 4. I don't see the conflict between conclusion 3 and premise 1 in this situation.
>
> Here you've perhaps identified the source of our disagreement. By picking some set of principles and some set of positions (even if these are fairly large sets) you're able to argue that those principles aren't useful for picking a move in those positions. I would agree. My counter-point was that by picking some other set of principles (just two endgame strategies) and some other set of positions (just one) I could argue that the principles (those that I picked) are useful (yes, there are only three drawing moves in that position, but the point of the principles is not a single move but a sequence of moves). Your response then is that those weren't the principles and positions you were talking about. That's fair, but then how do you conclude that higher-level players can never justify moves, or should not ask 'why'?
>
If I understand this part correctly it seems that this rests on the idea that the rook endgame example is an exception to the main argument of the post, although I cannot see why that is, given my clarification in the first part of this response.
@SmaragdElefant said in #39:
> This part is different to the other parts. First difference I didn't read the whole part. Second I can mostly agree.
My biggest 'fan' continues to read my posts (or, at least partially, even though you commented?) - thank you! :)
This is partly what I was referring to in part 1 when I was using the distinction between descriptive and practical to summarize my concerns (which, you gracefully called 'nonsense'). The descriptive perspective (set of rules, principles, guidelines, and more generally academic-type analysis) of chess is not particularly useful in influencing the practical side of play (the decision-making process).
Of course, I would not want to make these explicit claims in that part, as those without any patience would immediately start offering retorts before I had the chance to offer a comprehensive argument for this view, which is why I left that till now where I could dedicate a whole article to it.
>I don't know, if principals are made to evaluate moves. And therefore maybe the attack is on the wrong aim.
If that were the case, then I would have little problem with that view and the role of principles in aiding decision-making. However, unfortunately that's clearly not the predominant view within the chess world, which I tried to show using my examples at the start of the article.
>And I read an article about scientists try to understand how our brain works and studying chess grandmasters and try to >explain their results on playing chess. They said it is mostly like the quote of Capablanca: "I see only one move ahead, but it is >always the correct one."
Unlike most areas of chess, there is actually a decent amount of academic research in this area, since it relates to decision-making and intuition on a more general level (which is more likely to get funded and pursued than if the aim was improving the abilities of chess players). The conclusions of this research generally agree with what you have written and the conclusions of this post, although I haven't mentioned them explicitly yet. I'll be discussing more of the empirical side in the next part.
>Or they are similar to grammar rules according to the article. Just because you can speak a language, don't mean that you are >able to write down the grammar rules.
That's exactly right. Generally speaking, chess 'justifiers' are linguists, and chess masters are fluent native speakers. Of course, this analogy doesn't represent an argument itself, rather just an intuitive way to think about the main position/conclusion of the post.
@Wodjul said in #40:
Thank you for that comment! I really enjoyed reading it as you bring up several important points.
> Surely, the intent of general principles is to increase the probability of good human moves and decrease the probability of bad human moves.
Yes I think this is a common line of thinking and one that I've considered. Here's what I've concluded about this probabilistic conceptualization:
Chess has certain rules. These rules lead to certain game properties. These game properties lead to general principles. These principles, in some vague sense, are generally true.
For example, if you get checkmated, you lose. So protecting your king is important. Because a lot of the action in chess happens in the middle of the board (another property you could probably further 'deduce' if needed), it's best to hide your king in the corners early on. This might lead some principle saying that you should try to castle your king in the opening. We might think that this is generally true. Furthermore, even though very unlikely in reality, we could imagine ourselves doing some kind of perfect statistical analysis, and ultimately find that this is true 70% of the time (whatever that means...).
However, given this, how will this help us as chess players? To perform even at an amateur level in chess, it requires a nuanced, position-specific approach. This is why I believe chess players naturally develop the type of decision-making process I outlined in part 3, as it best solves this issue. No player closes their eyes, and simply trusts general 'external information' to make decisions in chess. Rather, we trust our specific intuitive appraisals in any given unknown position. This is why principles help boost confidence (as I described in the post), but never remove confidence. We would never see a move that immediately feels like it must be a great move, but then on consideration of the fact that, in general we should castle in the opening, we now feel that our move is not so good. We might change our mind upon calculating more, or spotting previously unseen moves, but not by simply recalling some general truth. The awareness of some general abstract 'truth' derived from first principles fails to influence our decision making given the highly specific nature of chess.
I don't think the above reasoning is necessary for supporting my general stance, as I believe my main argument in the article already does a good job of this by asking how exactly knowledge of probabilistic, general principles concretely influences an initial move preference. Nonetheless, I think the above is a good overview/perspective to understand why such general principles may not be useful for players deciding between a set of candidate moves.
>For beginners like me, it is good (I believe) to attempt to look for forcing moves first off; for the enemy (in case I need to >defend) and then for me in case there is a winning combination immediately available (like in a puzzle).
Indeed as you mentioned, I would consider such things as personal protocols rather than general, player-independent principles. These are mental tools that players may or may not equip themselves with to alter their decision-making process for better results. For example, since beginners often blunder, having some kind of explicitly defined habit, personal rule, or in general 'slow-thinking' process (blunder check, reminding oneself to look for checks and captures) can be useful. The same could of course be said for principles, although as I've argued, they cannot influence an initial move preference. I could also potentially argue against the utility of certain 'personal protocols', but then that would be a separate discussion.
> General rules (highly sophisticated and extensive ones at GM level and often internalised perhaps not
> even in standard language like English or Russian) do exist and make sense for humans (I contend).
This leads to an interesting discussion. If I'm understanding correctly, it focuses on the idea of the existence of certain principles. This relates to what I was discussing in part 2 where I discussed pragmatism in chess. Is it 'true' that we should generally castle early in the opening? In some sense, I guess it might be true. The truth of this statement might embody itself in the games of strong players.
Although, from a practical perspective, if it's not clear how this idea will help us play better, then the truth value of that statement is irrelevant. Are the rules of grammar, that seemingly manifest themselves in the language of native speakers, true? Well, if our goal is to become a fluent speaker, and if the rules of grammar cannot help us achieve our goal, then who cares if they are generally true, or if they seemingly dictate the language of native speakers.
>Yet, I view these theory essays by @DailyInsanity as very interesting exploratory essays in pure reason. These are "notes >towards" a theory but not a full theory yet. I can understand the impulse to explore this.
Thank you very much and I agree! While I still have a lot more to write, I agree that even by the 'end', it will not constitute a totally comprehensive theory or anything like that. I do, however, hope to start a more productive discussion in the chess community about chess improvement, which leads to this point:
>Finally, when using speculative pure reason for deductions (supported indeed by premises with some potential "truth warrant" >as the philosophers term it) there will be a need to develop, if possible, testable hypotheses.
I absolutely agree. This is related to what I was mentioning at the start of part 2, where I state that for any empirical aspect to have direction, we must develop a solid theoretical understanding in order to propose 'worthy' hypotheses to test. I'm planning on proposing such hypotheses, based on the conclusions made, in further posts.
@dboing said in #41:
> so we dissect certain things but not others. how about dissecting the word principle?
You're absolutely right that we could 'go deep' on this front too. There are actually many aspect where I intentionally haven't done so, mainly as to not 'dilute' my main argument, which I think most can understand without the need for an extra theoretical and pedantic prelude. Maybe in future, more extensive writings on the topic I will do so for completion and for personal satisfaction reasons.
Another example of what I omitted was the idea of 'applying' a principle. Even if we could somehow settle on a single principle in any given position, how do we exactly 'apply' that principle to determine a specific move? After all, a single principle can simultaneously suggest several distinct moves. This is another argument to be made here. However, again for the same reason of avoiding 'dilution', I omitted it from the post.
A brief note here for @DailyInsanity and anyone else who might be interested. I am going to post a topic in Off Topic Discussion. This will keep it separate from this topic so that I don't go off-topic here. It will also mean people only have to read and maybe post there if they are directly interested in the subject(s) I raise. It will be about philosophy generally and ontology specifically. In turn, the types of ontology I overview will involve only the ontologies of real systems and formal systems but NOT religious and speculative ontology (or pure metaphysics) which I will not go in to.
There is a point of intersection in that I have found computer games and computer simulations (of several types) raise or bring to the forefront very interesting ontological points about the interactions and intersections of real systems and formal systems. The concept of modelling (model systems, real and virtual and their interactions and relations with and to the more complex target real systems) will be involved also. I won't say any more here other than to say my game interests have included both RTS (Real Time Strategy) games and board games.
I suppose I have to weigh in, as I have some objections to some of the premises and conclusions made. This will be in two parts, the logic and then the example.
The Logic
Let me summarize the logic you laid out in my own way so as to check whether I understood the meat of your argument. In essence, you're saying this is like astrology. If one can predict anything using astrology, then astrology truly predicts nothing. Similarly, if principles can be used to justify anything, then these principles truly justify nothing.
Here, I believe, you've set up the argument that would lead to a false dichotomy. Either the principles are useful, and if not, they're useless. One Zero. The hypothesis is set up in such a way that if result A happens, then useful, if result not A, then useless.
Let me set up an alternative hypothesis, which will not be as susceptible to a false dichotomy.
Given a position with x possible moves, the moves which receives the most concurrence with several principles from p1 to pn and the least objections from other principles from p1 to pn would rise to the top.
In your example, you made it sound like b4 is an absolutely dumb move, but that is not the case. It is top 4 top 5 out of x moves.
I'm betting that as you go down the list into worse and worse moves, fewer principles would be applicable, and more principles would be violated.
What I'm suggesting is that use of principles is not an all or nothing prospect. Principles could be extremely useful, very useful, mildly useful, somewhat useful, etc. I'm also not discounting the fact that there's some qualitative unverbalized aspect controlling these principles. All I'm saying is that principle utility is not not one zero, and I'm leaning towards principles being very useful.
The Example
Allow me to make an analogy. Let's use stockfish parameters as our proxy for principles. Surely this is not a reach since these parameters were conceived with chess principles in mind. Even if we provide that some parameters indeed have larger weights (meta principles?, intuition?) than others doesn't discount the fact that these parameters (principles) do exist and are being utilized by the engine.
If we are to use this analogy for humans, we learn some set of principles and then we use our intuition to guide us which principles or combinations of principles are relevant for any given position. For example, if we select a move precisely because it would lead to us having a bishop pair and an open position means that these principles have utility.
Final Thoughts
Be that as it may, objections to some of your premises and conclusions aside, I actually don't have any objections to the move selection / learning approach you are proposing. Seems like what you're trying to do is replacing principles with intuition. This worked with engines (Lc0, A0, etc.), so why not try this out for us?
As I see it, there are two problems with this intuition approach to chess:
1. In an actual game, how exactly does one select candidate moves with an intuition only approach?
2. In teaching, how can concepts be taught to others without resorting to principles?
I assume you already have solutions for both, and I'm quite excited to see how that would look like.
slippery words....
@DailyInsanity said in #45:
> Another example of what I omitted was the idea of 'applying' a principle. Even if we could somehow settle on a single principle in any given position, how do we exactly 'apply' that principle to determine a specific move? After all, a single principle can simultaneously suggest several distinct moves. This is another argument to be made here. However, again for the same reason of avoiding 'dilution', I omitted it from the post.
Why the need for a single best move. If we are assuming a learning space (which has not been made explicit, but I do, as it helps me ramble less), with individual initial conditions of some chess skill map (not wanting to get bogged down, but i could elaborate) to be be evolved, so we could have some pragmatics there to. So we could start talking about the training set meaning full size w.r.t. larger set size. (which implies certain things that also allow positioning within those sets, and subset sizes as well for the specialist notions and the generalist notions to be part of the language of the model you could still build on top of what I think you are building so, far (blog 3, for me). I think the question of generalization does not have to stay a black box (I can also elaborate why I think that). Those are all related, in my set of ideas, I find myself exhausted trying to share intelligibly using the language of 3D printer filament walks in the 3D space, here it ain't 3D, but I might be using 3D slices in minds eye at times).
Has the single best move been the assumption all along in all the discussions? What should there be also, one line of argument for the same game? If finally doinjg the pragmatic thing of making the learning trajectory (it might not be the trajectory of a point, it might be the trajectory of a function or the evovling maps, if we have seeing and evaluating as different co-domains, a shape trajectory of sorts, I have dealt with such things in my past experience, non in chess, of course, unless we are willing to talk about contours on the board of certaing combination of things, one day).
The I think that we might want to relax the single best anything for a positon. and even the same ulimate goal. I might have been misreading or jumping from that reading chunk. I am sorry, if so. I have hard time with dense informative text, even when not written by me... which would explain it, but no. The same reason i may be hard to read, makes it hard for me to read without reacting, when the read is stimulating. And when it isn't well, I don,t even keep on it. This is one aspect of the nature of my exhaustion. I do like the topic here a lot. I have other limitations.. too. But that internal constraints has been with me for a long time. I just can't work around it, with the added other ones. So, I will come back here.. I find the others might be doing the same job I thought I could participate of giving you the helpful and necessary complementary points of view that a proper constructive but critical research project through discussion would need. Based on the probably axiomatic assumption that we are all fallible, and especially on the rigorous reasoning. We can't be trusted the same way a chess champion can be trusted to find the best move, about building theories that are actually veribifable or transmissible through the human mind solitude. Language is not that great to bridge such gap, when the pointed reality of convention and common sensory experinece is not a 3D printer filament itself.