A related topic in this interesting discussion are the compliments. Chessplayers are rarely congratulating each other so we lose a lot of opportunities to get positive feedback.
2 articles on my blog touched that topic:
schaken-brabo.blogspot.com/2023/02/complimenten.html
http://schaken-brabo.blogspot.com/2023/09/verrassingen-deel-3.html
2 articles on my blog touched that topic:
schaken-brabo.blogspot.com/2023/02/complimenten.html
http://schaken-brabo.blogspot.com/2023/09/verrassingen-deel-3.html
Another related topic is even things which are already good can sometimes still be improved. We need to be careful of course not to get dragged into insignificant details but at the same time don't get too lazy. It is something I discussed already in 2013 on my blog : http://chess-brabo.blogspot.com/2013/09/which-games-to-analyze.html
A third aspect is the "modern" philosophy: Work on your strengths, not your weaknesses. There are still debates ongoing in how far can this be stretched as there are obvious limitations. Anyway it is definitely not nonsense either.
Plug the leak and the container will hold water. So fix the weakness and you will be able to hold ground. Then specialize to excel in a domain. A player cannot be exceling in all domains, unless they've been fixing every weakness they have compared to their strengths.
If Lichess had an Rating graph like the Lucas Chess Elo average graph, we could discover something new about our games.
At least Lichess has a dashboard that shows our tactical weaknesses, compared to our other tactics. It helps to know where we excel and where we need to improve on. Only players that want to improve need to use it. lichess.org/training/dashboard/90/dashboard
If Lichess had an Rating graph like the Lucas Chess Elo average graph, we could discover something new about our games.
At least Lichess has a dashboard that shows our tactical weaknesses, compared to our other tactics. It helps to know where we excel and where we need to improve on. Only players that want to improve need to use it. lichess.org/training/dashboard/90/dashboard
yes it is possible to learn from won games as well, but it is likely to be less efficient.
First of all, by analyzing our losses and finding the exact point where we went wrong, we already implicitly learn that we played well up to that point. Hence, positive reinforcement is already included to an extent.
Secondly, if the opponent made an obvious blunder then there's not too much to learn from and if they didn't it's harder to attribute what moves contributed to the win than it is to attribute what own moves contributed to a loss.
If we do find the move(s) that contributed to the win and they are part of a combination or a specific strategy, then, unless we stumbled upon them by complete accident in the game, we already possess the skills and knowledge for said combination or strategy, so we can't learn anything new from it and the take away is restricted to a reinforcement of a concept that worked.
And lastly, unless we eliminate our weaknesses, the opponents will consciously target them. It is therefore more important to eliminate weaknesses than to further improve strengths. Weaknesses are more easily detected in losses.
In summary, there is a reason why the general advice is to learn from your losses and a random study from a very different (in my view not transferable) context is probably not going to change that.
First of all, by analyzing our losses and finding the exact point where we went wrong, we already implicitly learn that we played well up to that point. Hence, positive reinforcement is already included to an extent.
Secondly, if the opponent made an obvious blunder then there's not too much to learn from and if they didn't it's harder to attribute what moves contributed to the win than it is to attribute what own moves contributed to a loss.
If we do find the move(s) that contributed to the win and they are part of a combination or a specific strategy, then, unless we stumbled upon them by complete accident in the game, we already possess the skills and knowledge for said combination or strategy, so we can't learn anything new from it and the take away is restricted to a reinforcement of a concept that worked.
And lastly, unless we eliminate our weaknesses, the opponents will consciously target them. It is therefore more important to eliminate weaknesses than to further improve strengths. Weaknesses are more easily detected in losses.
In summary, there is a reason why the general advice is to learn from your losses and a random study from a very different (in my view not transferable) context is probably not going to change that.
<Comment deleted by user>
<Comment deleted by user>
Interesting
en.wikipedia.org/wiki/Reinforcement_learning
There is also the old and developed behaviorism approach, but I find it less explanatory than the well formulated machine simulation of the basic model of learning, such as in A0 and Lc0. In chess.
In behaviorism, for experimental reasons of the past, and wanting to avoid too much subjectivity in the buildup of the psychological theory, care was taken to have very "atomic" (that can't be dissected further) notions of behavior. They had to be measurable in short time from an external observer point of view only (I think it was a port of ethology to human behavior), it has its virtues of internal rigor, but might be swooping a lot of higher cognitive functions under the carpet.
Yes, it is very old in psychology models research. Past century old (well said like that could sound old to some generation). But I don't think that higher cognitive functions are going to change this ubiquitous fact that comes from an CNS wiring needing to adapt to some environment to keep on going. (or thrive). The thrive might be missing, as too subjective for that explanation framework, at the time. It was an attempt at putting some rigor in psychology research constructions. I think there is now a more balanced approach, with adjusted experimental methods that can include other end-point than pure external observables with short term behavior chunks (my words).
There is also the old and developed behaviorism approach, but I find it less explanatory than the well formulated machine simulation of the basic model of learning, such as in A0 and Lc0. In chess.
In behaviorism, for experimental reasons of the past, and wanting to avoid too much subjectivity in the buildup of the psychological theory, care was taken to have very "atomic" (that can't be dissected further) notions of behavior. They had to be measurable in short time from an external observer point of view only (I think it was a port of ethology to human behavior), it has its virtues of internal rigor, but might be swooping a lot of higher cognitive functions under the carpet.
Yes, it is very old in psychology models research. Past century old (well said like that could sound old to some generation). But I don't think that higher cognitive functions are going to change this ubiquitous fact that comes from an CNS wiring needing to adapt to some environment to keep on going. (or thrive). The thrive might be missing, as too subjective for that explanation framework, at the time. It was an attempt at putting some rigor in psychology research constructions. I think there is now a more balanced approach, with adjusted experimental methods that can include other end-point than pure external observables with short term behavior chunks (my words).