I really like your introductions and i should say that i have too many distracting thoughts while i read interesting things, that i rarely keep going, if i don't get some of those out first, but all of that takes some time, and i don't get to finish reading or doing much,). So some of that.
> The Random Skills of Adult Improvers
Rewriting for effect. Ok... Random is a buzz word in chess. Good device. But so is the tyranny of title choices, that it might make a passer by miss the boat, because of so few words, that, indeed, it is good device.
> One of the weirdest things I have noticed about Adult Improvers [...] is that we have a bizarre skill distribution in chess.
Ok i will let "bizzarre" pass as a transition from random... (I am actually more humble, and proud of that, than that sounds, and i might also contradict myself, at times, etc...).
I would say, not random (or define random or as opposed to what). Maybe: not expected? I would also say, no, it should be expected out of a complex thing to learn about, which is still mostly an art form, to master.
I would say that if one were to give some credit to machine learning as model of learning (it actually has had many years of cohabitation with actual cognitive psychology research over recent decades)), one would expected a strong association between experience set and the characteristics of that set with the learner self-confidence, or sense of confort in planning (any range), and probably actually "regional" levels of "accuracy".
I suspect you would agree.... But it popped out loudly while reading.... I am learning about how to bring about topics with some life in the text from reading your introductions....
Here is hoping i can read further, as i also would like to see your supporting points for my own later use as knowledge sharing material.
> The Random Skills of Adult Improvers
Rewriting for effect. Ok... Random is a buzz word in chess. Good device. But so is the tyranny of title choices, that it might make a passer by miss the boat, because of so few words, that, indeed, it is good device.
> One of the weirdest things I have noticed about Adult Improvers [...] is that we have a bizarre skill distribution in chess.
Ok i will let "bizzarre" pass as a transition from random... (I am actually more humble, and proud of that, than that sounds, and i might also contradict myself, at times, etc...).
I would say, not random (or define random or as opposed to what). Maybe: not expected? I would also say, no, it should be expected out of a complex thing to learn about, which is still mostly an art form, to master.
I would say that if one were to give some credit to machine learning as model of learning (it actually has had many years of cohabitation with actual cognitive psychology research over recent decades)), one would expected a strong association between experience set and the characteristics of that set with the learner self-confidence, or sense of confort in planning (any range), and probably actually "regional" levels of "accuracy".
I suspect you would agree.... But it popped out loudly while reading.... I am learning about how to bring about topics with some life in the text from reading your introductions....
Here is hoping i can read further, as i also would like to see your supporting points for my own later use as knowledge sharing material.
Great post Ono, really enjoyed it!
That was a cool post! I never thought of it that way. I would say my main strength is tactics. If i focus, I spot tactics people my rating don't. Endgames are a huge weakness though. I need to buckle down and work on endgames, my rating would soar.
Interesting point about how there's no "rook endgame" rating. (Although technically wrong.)
The concept of mini-games is absolutely right tho, for sure. Every single position is a thing, and to some extent we can cluster them into categories.
-- frentz
The concept of mini-games is absolutely right tho, for sure. Every single position is a thing, and to some extent we can cluster them into categories.
-- frentz
I think chess games should start from any position consensually agreed to. sort of 960 for the standard masses...
While phase may be a region of experience set, I would think that more position based characteristics might also correspond to experience set correlated abilities.... one might even end up with some sets of patterns, as a good way to separate skills, trigger patterns and dynamic patterns associated...
My previous post referring to machine learning, was more in that light.. But, yes endgames, material attrition is likely to make challenges that have less distractions or noise from gone material, allowing minimal pattern examples to stand out...
There is a mini-chess game, that I recently saw as the topic of a A0 framework application, same formalism, and mathematics of solution, but less complexity.. However, it also, seemed to consider only one initial position, and full games from there (i need to read it actually, abstracts don,t always tell the gist of the content). Not the same thing as here, as a combinatorial complexity reduction transition set up.. Why not use standard full board and just consider all the mini games, tying them to themes and pattersn.. starting from smallest necessary amount of material, usual endgame.
Caricatures are not just fun, they are telling. So that when an adult improver, which is usually more self-aware and able to guide one self through the experience set in relation to the results, and can hone certain aspects better than a child that way, ....
wait, the sentence is not complete:
... so when that the above phrase, and the adult starts, not building an opening repertoire, but a familiarity with characterized mini-games (in the spirit of above), which they may not have words for, but still do the necessary self-guidance while playing to keep exercising around the skills/pattern they are smelling, ... (still not complete the sentence)...
... then I think endgame is likely to be the smoothest place to start with and expand from backward..
unfortunately, there is no glory there.. and chess is taught bacward all the time, from the most complex position of them all... the standard initial one.. Even A0 and LC0, reinforcement learning algorithm, is stuck that way.. (i claim that maybe that is why it has more resolution in openings, and less in endgames).... so it might learn the hard way, in spite of gettings thrown in the deep end of the pool. and so are we..
While phase may be a region of experience set, I would think that more position based characteristics might also correspond to experience set correlated abilities.... one might even end up with some sets of patterns, as a good way to separate skills, trigger patterns and dynamic patterns associated...
My previous post referring to machine learning, was more in that light.. But, yes endgames, material attrition is likely to make challenges that have less distractions or noise from gone material, allowing minimal pattern examples to stand out...
There is a mini-chess game, that I recently saw as the topic of a A0 framework application, same formalism, and mathematics of solution, but less complexity.. However, it also, seemed to consider only one initial position, and full games from there (i need to read it actually, abstracts don,t always tell the gist of the content). Not the same thing as here, as a combinatorial complexity reduction transition set up.. Why not use standard full board and just consider all the mini games, tying them to themes and pattersn.. starting from smallest necessary amount of material, usual endgame.
Caricatures are not just fun, they are telling. So that when an adult improver, which is usually more self-aware and able to guide one self through the experience set in relation to the results, and can hone certain aspects better than a child that way, ....
wait, the sentence is not complete:
... so when that the above phrase, and the adult starts, not building an opening repertoire, but a familiarity with characterized mini-games (in the spirit of above), which they may not have words for, but still do the necessary self-guidance while playing to keep exercising around the skills/pattern they are smelling, ... (still not complete the sentence)...
... then I think endgame is likely to be the smoothest place to start with and expand from backward..
unfortunately, there is no glory there.. and chess is taught bacward all the time, from the most complex position of them all... the standard initial one.. Even A0 and LC0, reinforcement learning algorithm, is stuck that way.. (i claim that maybe that is why it has more resolution in openings, and less in endgames).... so it might learn the hard way, in spite of gettings thrown in the deep end of the pool. and so are we..
Hey, I really think it would be a great idea to start split ratings informatically for futureLichess puzzels like on a parental tree scheme and then calculate the average of the "mini games" like you say partial in a therotical rating sum. It would be more sound than the puzzle rating on lichess right now. But besides unavoidable inflation, a limited pool of players etc. : No engine and no one can catch up the psychological point though, if for example any opponent uncover her/his weakness with an obvious questionable move, any practical player will go for it anyway without doubt, and do the fake magic to win, not to mention the flaggin stuff, or artificial theretical draw prolongiation in games limited by the clock. Separating theoratical chess maturity from practical chess maturity is another topic anyway, but let it being said that your idea hold didactic GOLDright now.
I agree
I agree
@PartyMcFly said in #8:
>
Actually you might want to visit the puzzle dashboard. Although it lacks the hover feature sharing the sample size, it does display "performance" rating per lichess tactical patterns (or theme if distinction have to be made).
So if the sampling was big enough and the theme was crucial to the hard thinking effort in solving the puzzle, not just happening to be there (some themes are not really patterns to thingify on the board, like mate in 2, short puzzle, or level of advantage), then we would have something very close in essence as a basis to decompose the average puzzle rating into components (even if not really independent or orthogonal), strenghts and weaknesses might be induced from those..
That is a good foundation, for what I think you are proposing (we are all, the blog author as well, if not mistaken, this is on topic). In the tactical skill business.
I agree, too, and I have been very interested by that innovative aspect of lichess puzzle system, since version 2.
I also like the population feedback system (wherever it lives as a database) about them.
I like to learn the themes definitions and how they can be stretched given experienced puzzle challenges, it is a bidirectional learning effort, test the words, but also adjust the chess position information pointed by the words, i mean figure out the chess things to associate with the worded patterns.
in brief.. check all the puzzle derived pages when you finnish a series of puzzles.. do a few rows.. within 90 days.. this is the limit of the statistics compilation in terms of components.. (it could be number of puzzles, but hey things are what they are, said the river).
>
Actually you might want to visit the puzzle dashboard. Although it lacks the hover feature sharing the sample size, it does display "performance" rating per lichess tactical patterns (or theme if distinction have to be made).
So if the sampling was big enough and the theme was crucial to the hard thinking effort in solving the puzzle, not just happening to be there (some themes are not really patterns to thingify on the board, like mate in 2, short puzzle, or level of advantage), then we would have something very close in essence as a basis to decompose the average puzzle rating into components (even if not really independent or orthogonal), strenghts and weaknesses might be induced from those..
That is a good foundation, for what I think you are proposing (we are all, the blog author as well, if not mistaken, this is on topic). In the tactical skill business.
I agree, too, and I have been very interested by that innovative aspect of lichess puzzle system, since version 2.
I also like the population feedback system (wherever it lives as a database) about them.
I like to learn the themes definitions and how they can be stretched given experienced puzzle challenges, it is a bidirectional learning effort, test the words, but also adjust the chess position information pointed by the words, i mean figure out the chess things to associate with the worded patterns.
in brief.. check all the puzzle derived pages when you finnish a series of puzzles.. do a few rows.. within 90 days.. this is the limit of the statistics compilation in terms of components.. (it could be number of puzzles, but hey things are what they are, said the river).