Comments on https://lichess.org/@/nanajoon/blog/why-engine-analysis-doesnt-actually-teach-you-chess/KJ3IXNa9
This is an ad for ChessLogix. That is OK and allowed, but should be declared.
you know, lichess is a free and open source project. so when someone creating a project related to lichess or advertising in lichess, people expect that product to be free and open source too .
@tpr said in #2:
> This is an ad for ChessLogix. That is OK and allowed, but should be declared.
I don’t have any affiliation with ChessLogix, as I noted in the blog post. I’m just a user sharing a tool I personally found useful for post-game analysis. I’d been looking for something like this for a while, and I genuinely find it valuable.
They also asked me for feedback, and they actually implemented a few of the features I suggested. That responsiveness was a big part of why I felt it deserved a shout-out, and what motivated me to write the post.
@aminson123 said in #3:
> you know, lichess is a free and open source project. so when someone creating a project related to lichess or advertising in lichess, people expect that product to be free and open source too .
That’s a totally fair expectation, and it’s one of the reasons Lichess is special.
I shared this because I genuinely found it useful, not because I think it follows the same free/open-source philosophy as Lichess. Also, before posting I checked the Lichess blog guidelines and made sure this post stays within them.
Out of curiosity, did you find the content of the post itself interesting, at least the first part discussing engine-based game analysis and why it often falls short in helping players recognize their recurring bad patterns?
@Nanajoon
The title is: *Why Engine Analysis Doesn’t Actually Teach You Chess.*
Did anyone claim that tho? Of course engine analysis doesn't teach you chess in itself.
The reason you wrote that obvious statement was so you could swoop in with saying that the AI chess website your advertising actually helps (*compared to staring at engine analysis*).
It's not exactly a show of support for a website that you require it to be measured against the most obviously unhelpful way of chess training.
That AI thumbnail is bad vibes too.
*There's also a strange incoherency:*
So in the image it says 'Stockfish engine paralyzed by infinite analysis loops'.
The reason thats nonsensical is because doesn't become 'paralyzed'. It gives you the eval in seconds. Infinite analysis means the engine increases the depth, but that doesn't paralyze the engine or prevent an eval being given.
It's just a nonsensical thing to say.
There seems to be a correlation with AI thumbnails and incoherent text being written on the image.
>There’s a subtle but powerful shift when a mistake is given a name.
“Blunder” is vague.
“Phase Transition Blunder” is specific.
“Bad move” is forgettable.
“Central Push Blindspot” sticks.
You must be joking. 'Phase Transition Blunder' is incredibly vague. So is 'Central Push Blindspot'.
The AI advice is hollow and superficial. Phrases like 'harmonized pieces', 'centralize' and 'consolidating the center' are used to try to hide this.
@RuyLopez1000
I appreciate you reading the post and engaging critically with it. I recognized your name because I read your blog, “AI Slop is Invading the Chess World,” and all the discussions and comments there, in particular your heated discussion with one of the Lichess developers. First of all, I really enjoyed reading your blog post, and I agree with most of what you discussed in that post, mostly because I have personally tried and tested many superficial AI products with bold claims that they would become your chess coach. And as you pointed out, and many others did too, from the first game analysis you can quickly spot how ineffective most of these tools are. I do not want to get into that discussion, but just to say that I totally agree with @totalnoob69 's comments on your post that AI is just a tool (a great one), but it is nowhere near human intelligence or Artificial General Intelligence (AGI), as many “frontiers” of current LLMs like Sam Altman have claimed it to be, or soon to be. There have been numerous recent papers and articles discussing the fundamental limitations of Transformer architectures (like limited context memory or limitations of chain-of-thought reasoning) that stop the advancement of GPT-like models from achieving AGI. That being said, a lot of recent AI products have been initiated and developed with that hype and those high expectations, and that’s the main reason a lot of these tools fail. You, in your blog post, pointed out some of the flaws that are mostly about how superficial these tools are in terms of reasoning about a chess position or game. I totally agree with your points about that. But I must also say that just because AGI is not here, it does not mean the current LLMs and GPT models are completely useless in terms of making improvements to the engine analysis we currently use. LLMs are not as smart as humans, but they are still definitely one of the most impactful inventions of this century (though obviously not every product that stems from it is). So, there is a flaw in an argument that I hear a lot, where people completely dismiss LLM capabilities simply because they are not as smart as humans.
Now back to the points you mentioned in your comment here.
> "The title is: Why Engine Analysis Doesn’t Actually Teach You Chess.
Did anyone claim that tho? Of course engine analysis doesn't teach you chess in itself."
I do not know about you, but in the world of chess everyone does that. Especially if you lose a game and you are not sure where and how the game flipped in your opponent’s favor, the first thing people do is check the engine analysis to understand that. Even GMs do that sometimes to analyze their future opponents to study and understand possible flaws and potential lines to challenge them. All players have blind spots or bad habits, and you either need someone like a coach to review your games, or you do it yourself using engine analysis. Not sure why you are surprised by that claim.
> "The reason you wrote that obvious statement was so you could swoop in with saying that the AI chess website your advertising actually helps (compared to staring at engine analysis).
It's not exactly a show of support for a website that you require it to be measured against the most obviously unhelpful way of chess training."
There are multiple reasons why I wrote this blog post:
I have been looking for a tool like this for a long time, and I have tried many of them. Every time I watch a game analysis on YouTube by Levy Rozman (GothamChess) or Agadmator, I wonder how great it would be if someone could do that for other interesting games by top players. So it’s been a while since I have been thinking about this and looking for something that explains engine lines, or explains what’s wrong with a move (from a positional point of view, not just tactical).
I am suffering from a rating plateau. Even though I study openings and do a lot of post-game engine analysis, I am still struggling to even maintain my blitz rating, and I really do not understand why. And I have read about this issue in many forums, including Reddit and here too.
And because of that pain point, I try every tool I find (at least give it a shot). One of them was ChessLogix, along with others like Sensei, Chessvia.ai, and Noctie.ai. And based on my experience, ChessLogix is by far the most promising one. Now do not get me wrong, I am not saying it would be as effective as having a real coach review your games and instruct you but it has been effective in terms of shedding light on blind spots and bad habits. And so far it has worked for me. I also gave some feedback to the developers through their website, and they contacted me and actually opened a discussion to implement and add the features that I proposed based on my experience using their website. The fact that they were open to taking user suggestions, and cared enough to turn their product into what users want, is what made me mention them here.
> "So in the image it says 'Stockfish engine paralyzed by infinite analysis loops'.
The reason thats nonsensical is because doesn't become 'paralyzed'. It gives you the eval in seconds. Infinite analysis means the engine increases the depth, but that doesn't paralyze the engine or prevent an eval being given."
The subject in that sentence is not Stockfish. It is me, as a user of Stockfish, who is paralyzed by the depth of the engine. The reason for that paralysis is that I do not understand the “why,” and it does not help me understand my mistakes and hence the rating plateau. I did not realize it would be confusing. Sorry about that.
> "You must be joking. 'Phase Transition Blunder' is incredibly vague. So is 'Central Push Blindspot'."
I think the argument is treating this as binary when it isn’t and I don’t buy the all-or-nothing framing. Consider “Blunder” versus “Central Push Blindspot.” Obviously, one has more detail than the other, and that helps you remember things better. Also, this was not the main explanation, it was a phrase or title as the root cause of why I lost that game. In fact, this was one of my main blind spots: I developed a habit of overextension with my central pawns because I am usually afraid of opening the center in blitz, where you do not have much time to think and you want to keep the position simple. I realized that blind spot after using this tool, and yes, the phrase “Central Push Blindspot” was very useful for me to say out loud whenever I was tempted to follow my habit of closing the center with overextended pawn pushes.
> "The AI advice is hollow and superficial. Phrases like 'harmonized pieces', 'centralize' and 'consolidating the center' are used to try to hide this."
Can you point out how you came to that conclusion with an example? I mean, I do not think this is the perfect tool I have ever seen, but it is useful, it gives good explanations, and it opens your perspective on chess understanding by highlighting and explaining mistakes from your games. So, it is not perfect, but definitely useful.
Again, I appreciate you taking the time to read this post and write the comment, and I hope we can have a constructive discussion. I’m open to strong criticism, but I think this conversation would be more productive if we kept it focused on the substance rather than mockery.
The subject in that sentence is not Stockfish. It is me, as a user of Stockfish, who is paralyzed by the depth of the engine.
Then why does it say 'Stockfish engine paralyzed by infinite analysis loops' and not 'I'm paralyzed by infinite analysis loops'?
And why is there an image of a loading bar in a computer with a sad face? (Implying that Stockfish itself is stuck)?
Clearly you were saying that Stockfish get paralyzed by infinite analysis loops, not you.
"The AI advice is hollow and superficial. Phrases like 'harmonized pieces', 'centralize' and 'consolidating the center' are used to try to hide this."
Can you point out how you came to that conclusion with an example?
Many examples. Here are two examples from your screenshots.
So it says that basically you blundered (in a position where the pawn structure changed) and that led to tactics. You could figure that out yourself. The second sentence basically says 'you made a mistake with Qd4'. You don't need AI to tell you that.
'centralizing' - now you know that d4 is a central square
'without fully resolving forcing replies' - now you know that not 'fully resolving forcing replies' leads to blunders.
'make a consolidating move that keeps central tension while covering tactical shots'.
The AI hits all the right notes - consolidation, central tension, covering tactics etc. Basically you need to just play good chess lol.
No real details, just fluffy words.
You look through all the examples and you'll see the same pattern.
EDIT: You see the problem is with people who try to market this stuff. They wrongly make beginners believe that its useful for them. ChessLogix is trying to make a profit out of it.
If it's free then fine. But if they trying to make a profit out of this shoddy product through rhetoric then it's bad.
@Nanajoon
>The subject in that sentence is not Stockfish. It is me, as a user of Stockfish, who is paralyzed by the depth of the engine.
Then why does it say '*Stockfish engine* paralyzed by infinite analysis loops' and not '*I'm* paralyzed by infinite analysis loops'?
And why is there an image of a loading bar in a computer with a sad face? (Implying that *Stockfish* itself is stuck)?
Clearly you were saying that *Stockfish* get paralyzed by infinite analysis loops, not you.
>>"The AI advice is hollow and superficial. Phrases like 'harmonized pieces', 'centralize' and 'consolidating the center' are used to try to hide this."
>Can you point out how you came to that conclusion with an example?
Many examples. Here are two examples from your screenshots.
So it says that basically you blundered (in a position where the pawn structure changed) and that led to tactics. You could figure that out yourself. The second sentence basically says 'you made a mistake with Qd4'. You don't need AI to tell you that.
'centralizing' - now you know that d4 is a central square
'without fully resolving forcing replies' - now you know that not 'fully resolving forcing replies' leads to blunders.
'make a consolidating move that keeps central tension while covering tactical shots'.
The AI hits all the right notes - consolidation, central tension, covering tactics etc. Basically you need to just play good chess lol.
No real details, just fluffy words.
You look through all the examples and you'll see the same pattern.
EDIT: You see the problem is with people who try to market this stuff. They wrongly make beginners believe that its useful for them. ChessLogix is trying to make a profit out of it.
If it's free then fine. But if they trying to make a profit out of this shoddy product through rhetoric then it's bad.
@RuyLopez1000
I’ve already answered your question about the image and my intention. I made the image thinking the message was clear and a bit playful. I didn’t spend hours trying to optimize it for every possible interpretation beyond what I intended.
> So it says that basically you blundered (in a position where the pawn structure changed) and that led to tactics. You could figure that out yourself. The second sentence basically says “you made a mistake with Qd4”. You don’t need AI to tell you that.
>
> “centralizing” – now you know that d4 is a central square
That’s a very weak argument, so I’ll state the obvious. The point is not identifying which squares are central. The point is that my habit of trying to take control of the center with pawn pushes, without sufficiently considering tactics, is what caused the blunder. Later in the same game, I repeated the same pattern by centralizing my queen under time pressure.
Yes, I could identify the blunder with an engine, but the positional lesson would likely be lost there. By the way, the position I shared is quite interesting: there are multiple tactical continuations depending on how White responds if Black plays ...Nxd4. Also, my pawn push e5 was a very anti-positional move. So yes, the point is that with engine analysis alone, I would not naturally reflect on the underlying habit that repeatedly puts me under pressure and makes blunders more likely.
> “make a consolidating move that keeps central tension while covering tactical shots.”
>
> The AI hits all the right notes – consolidation, central tension, covering tactics, etc. Basically you just need to play good chess lol.
Yeah and learning through the right phrasing that makes you reflect on chess concepts is exactly the point. The goal is to understand what “playing good chess” means positionally and to describe it in a way that sticks. In many cases, this tool has done that reasonably well. Again, I’m not saying this is as good as having a real coach analyze the game with you, but it is definitely more helpful than engine analysis alone. That’s the main point I’m trying to make.
> If it’s free then fine. But if they’re trying to make a profit out of this shoddy product through rhetoric then it’s bad.
This sounds more like a criticism of capitalism than of the product itself. The argument seems to be: if it’s for profit, it’s evil; if it’s non-profit, it’s fine no matter how bad it is. But the more relevant question is about the product itself. Either it’s useful or it’s a waste of time. If it’s useful, then it may provide value to some people, and they can decide whether that value is worth paying for. If it’s a waste of time, then it’s bad regardless of whether it’s free or paid.
What I’m arguing is simply that I found this tool useful to some degree. It’s definitely not a replacement for a coach, but for me, it’s better than relying on engine analysis alone.
@RuyLopez1000
I’ve also written another blog post that specifically explains why it isn’t feasible for every project to be free, in response to your point implying that this tool must be free to be acceptable.
https://lichess.org/@/Nanajoon/blog/why-lichess-works-when-most-donation-models-dont/Fm90qcbx