Brilliants moves (!!) and others (Good, Book, Dubious) for game analysis

I wonder if the original post was written with AI.

@m011235 said ^

I wonder if the original post was written with AI.

What does this question has to do with the actual point of the post?

I have written it and used AI to give it a clean pass, to better summarize the ideas, and because I'm not a native speaker so I can clearly express my ideas.

@Axaet said ^

Your answer is inmature and doesn't add any valueable bit to the proposed idea.

Now I want to weigh in, too. You’re asking for an extremely tricky feature. All you’re contributing yourself are a few colored icons. And you expect the developers to handle the rest: not rocket science at all.

Next, someone will ask for a machine that turns lead into gold, suggest a few colors for the machine’s buttons, and then pressure the engineers to finally figure out the chemistry: not rocket science at all.

It's a relevant question. This is a forum for human discussion, what would be the point of a string of questions and replies by AI?

Your use case is valid, but instead of downvoting and aggressively asnwering to a valid question, you should have added a disclaimer note "partly written with AI".

Have great day.

@BauerA1 said ^

Your answer is inmature and doesn't add any valueable bit to the proposed idea.

Now I want to weigh in, too. You’re asking for an extremely tricky feature. All you’re contributing yourself are a few colored icons. And you expect the developers to handle the rest: not rocket science at all.

Next, someone will ask for a machine that turns lead into ...

I don't think that it is rocket science.

I'm a designer and partly a programmer. I know what I'm talking about. I develop using Python. I just don't use the language used by the Lichess developers.

It actually is not that hard. There is even no need to invent them. There are many different methods used.

As said, a GM, or a few, that knows about programming, data science, can easily build a robust system.
That is exactly why you have the (??) and the (?) already... and the whole Stockfish thing which developing that is far more difficult that what is being proposed.

Evaluating a game position to determine it is Brilliant, Good, or Book, is definetely no that hard and critical as rocket science is. As said, may it be difficult, yes, but there are more difficult things being solved.

The phrase rocket science was used to simply and contrast the argument to some saying "it is impossible to figure this out"... which extremely far from the truth.

I don't understand the bitter and stubborn air that the forum has...

@m011235 said ^

It's a relevant question. This is a forum for human discussion, what would be the point of a string of questions and replies by AI?

Your use case is valid, but instead of downvoting and aggressively asnwering to a valid question, you should have added a disclaimer note "partly written with AI".

Have great day.

Do you end your phrases were you learn english? Do you quote your english study books? I don't see that in your posts.

Guys, seriously, I imagined a more fruitful and interesting discussion in a chess forum... And something that adds something to the actual point of it.

Your comment is contributing anything to the original proposed idea.
If you are not going to add anything valuable, please just move on.

For chess move classifications (glyphs), if blunder / error / inaccuracy are essentially centipawn-loss (CP) derived, then satisfactory / good / brilliant should arguably be derived from a different dimension rather than simply being the positive mirror of CP. A move that loses little or no CP is already objectively “good,” so CP alone doesn't explain why some moves are considered especially strong or brilliant. Perhaps the missing dimension is computational difficulty: how much time, depth, or selective depth the engine needs to establish that a move is actually sound.

This makes depth vs. seldepth an interesting candidate. A small seldepth−depth gap could correspond to an ordinary, low-complexity satisfactory move; a moderate gap could indicate a good move that survives deeper tactical verification; and a large seldepth spike, particularly when the evaluation changes substantially as the search deepens, could identify moves whose strength depends on a long or forcing tactical horizon, the kind of move that initially looks questionable but is ultimately validated. In other words, the negative classifications measure how much evaluation was lost, while the positive classifications could measure how hard the engine had to search to establish the move's strength. The interesting question is whether a depth/seldepth-derived metric, perhaps combined with evaluation stability across depths, could provide meaningful thresholds for distinguishing okay, good, and brilliant.

@Toscani said ^

For chess move classifications (glyphs), if blunder / error / inaccuracy are essentially centipawn-loss (CP) derived, then satisfactory / good / brilliant should arguably be derived from a different dimension rather than simply being the positive mirror of CP. A move that loses little or no CP is already objectively “good,” so CP alone doesn't explain why some moves are considered especially stron...

Finally! This is the kind of replies I was hoping to happen.
Thank you for contributing on the subject.

Depth analysis and a contrast of them seems to be a good way to identify good, and brilliant moves.

@tpr said ^

"there is no simple algorithm for defining what is brilliant"

• There is: a move that:

1. wins,
2. is the only move that wins,
3. involves a sacrifice, and
4. is a quiet move, no check (+) or capture (x).
   It is easy to code.

"there is no generally agreed definitio what is blunder"

• A blunder (??) is a move that turns a winning position into a losing position.
  A mistake (?) is a move that turn...

I wish your definition of a blunder would go along with Lichess...
But whenever I lost “forced checkmate sequence” and still in a way better position, Lichess would give that move as a blunder rather an inaccuracy or something...