Well yeah, the gadget is 3500+. Of course we all have some trouble understanding its ideas (if we're being at all honest).

Not an expert, and somebody else would definitely give better explanation. I am just giving some superficial thoughts.

Example 1: 13. a5 would almost permanently shut black's light bishop out. For the same reason with 13. ... Bb7, black missed the opportunity to play a5 to relieve the bishop.

Example 2: 6. Bf4 is more difficult for me to criticise. But I would try to develop as black instead of the follow up Qa5+. On 6. Bf4 computer suggests 6. ... e6 which frees the bishop and taking the c7 pawn will be bad, because of Bb4+.

I listen to it as long as it's useful for my improving, as soon as it goes into endless lines - good bye, come back to the sea my old !

@kajalmaya said in #13:
> Not an expert, and somebody else would definitely give better explanation.
> I am just giving some superficial thoughts. ... 13. a5 would almost
> permanently shut black's light bishop out. For the same reason with
> 13. ... Bb7, black missed the opportunity to play a5 to relieve the bishop.
Indeed, after 1 d4 Nf6 2 c4 e6 3 Nc3 d5 4 Bg5 Be7 5 Nf3 O-O 6 e3 h6 7 Bh4 a6 8 a4 Nbd7 9 c5 c6 10 b4 Qc7 11 Bg3 Qd8 12 Bd3 a6 13 O-O a5, the machine seems to foresee increased activity for the c8-bishop after something like 14 b5 Bb7 15 bxc6 Bxc6 ...

In summary, Stockfish's commentary reflects the missed opportunities by White to either capitalize on their advantageous position or make the best defensive moves. The key themes in White's errors were the failure to challenge Black's pawn structure when advantageous, some untimely kingside aggression that weakened their own king's safety, and suboptimal piece maneuvering during critical moments.

re: lichess.org/JvIPFYj3#68

chat.openai.com/share/5e64a071-9db8-4bd3-b649-a6b80c14eef3

ChatGPT is more than capable of analyzing PGN and providing feedback. Sometimes it's super generic HOWEVER you can certainly ask it to elaborate and push it a bit more.

What's fortunate and somewhat surprising about chess is A LOT has been written about chess - although more of it in books than online - so it is able to use many games and analysis in its training data.

So yes I'm sure Stockfish can be a challenge to understand. Thankfully there's a superhuman AI assistant who may be able to provide guidance. That being said, I have a GPT-4 subscription which is way more powerful than 3.5 for doing this sort of thing...

@xDoubledragon said in #6:
> [Do you struggle to sometimes understand stockfish positional ideas? ...
> "... 1 d4 d5 2 c4 Bf5 3 cxd5 Bxb1 4 Rxb1 Qxd5 5 Nf3 Qxa2 6 Bf4 (1.76->0.00) Blunder.
> 6 Bd2 was best. (6 Bd2 Nf6 7 Qc2 Qd5 8 Qxc7 Nbd7 9 e3 e6 10 Bd3 Be7) ...
> After 6 Bf4, 6...e6 was best. (6...e6 7 Bd2 Nf6 8 e3 Nc6 9 Ra1 Qd5 10 Qa4 Bd6 11 Bc4) ..."]
When I try it, the machine does not seem to see that much difference between the two possibilities. After 6 Bd2, the machine anticipates Black holding White's advantage down to ~.6 after 6...Nf6 7 Qc2 Qd5 8 Qxc7 Nbd7 9 e3 e6 10 Bd3 Ne4.

Of course, it’s a machine. No humanity , it doesn’t feel, it doesn’t think, it has no psychology, bit like the terminator lol

@thefrickouttaherelol said in #16:
> chat.openai.com/share/5e64a071-9db8-4bd3-b649-a6b80c14eef3
>
> ChatGPT is more than capable of analyzing PGN and providing feedback. Sometimes it's super generic HOWEVER you can certainly ask it to elaborate and push it a bit more.

As usual, ChatGPT spits out complete nonsense. This time it's a bit more coherent, because it's basically restating what Stockfish said in addition to its own made up analysis which may or may not (usually not) make sense, so the nonsense is diluted by Stockfish's analysis. But the only information it adds is completely made up and often nonsense, and it's much more useful to simply look at Stockfish's analysis directly.

Please don't use ChatGPT to help you analyze games if you hope to improve. It will only hurt your improvement.

@AsDaGo said in #19:
> Please don't use ChatGPT to help you analyze games if you hope to improve. It will only hurt your improvement.

If it's nonsense then you'll ignore it or ask it again because it's nonsense. The point is to have a personal assistant help you see things that you normally wouldn't. Stockfish analysis is "shallow" when it comes to explaining its line of reasoning, even though the suggested moves are far better than any human could come up with.

The only thing that matters is whether or not you feel ChatGPT's analysis / explanations make sense. If so, then you have learned something and have benefited from it.