@FinalSpot If you want to improve the AI explanations, take a look at this project by @noobmasterplayer123
https://github.com/jalpp/chessagine-mcp
@FinalSpot If you want to improve the AI explanations, take a look at this project by @noobmasterplayer123
https://github.com/jalpp/chessagine-mcp
I don't have time, but if anybody wants to use Claude and the MCP above to auto-generate the rationales for the moves then I would be happy to donate some Claude credits and/or compute time /cc @FinalSpot.
Just PM me. You'd need to open source them and let me add them to the repository though.
I don't have time, but if anybody wants to use Claude and the MCP above to auto-generate the rationales for the moves then I would be happy to donate some Claude credits and/or compute time /cc @FinalSpot.
Just PM me. You'd need to open source them and let me add them to the repository though.
@Toadofsky said ^
OK, I'll ask: how did people select these puzzles?
I mined part of the Lichess database to look for puzzles that met a number of criteria (eg single winning move with no immediate tactical opportunities) and then asked 3 really strong players to go through them and filter out only the ones that were interesting to humans and to categorise them. So initially with code, and then with humans.
@Toadofsky said [^](/forum/redirect/post/EogyEink)
> OK, I'll ask: how did people select these puzzles?
I mined part of the Lichess database to look for puzzles that met a number of criteria (eg single winning move with no immediate tactical opportunities) and then asked 3 really strong players to go through them and filter out only the ones that were interesting to humans and to categorise them. So initially with code, and then with humans.
Thank you for doing this project @neilgd72
Thank you for doing this project @neilgd72
I don't know how to link to an individual puzzle but for this one, if you let Stockfish think long enough, the evaluation drops to 0 even with the best play:
https://lichess.org/analysis/standard/6k1%2F7p%2F1p4p1%2Fp1pp4%2F4r3%2F2PKP1P1%2FPP4P1%2F5R2_b_-_-_1_30
I don't know how to link to an individual puzzle but for this one, if you let Stockfish think long enough, the evaluation drops to 0 even with the best play:
https://lichess.org/analysis/standard/6k1%2F7p%2F1p4p1%2Fp1pp4%2F4r3%2F2PKP1P1%2FPP4P1%2F5R2_b_-_-_1_30
@AnlamK said ^
I don't know how to link to an individual puzzle but for this one, if you let Stockfish think long enough, the evaluation drops to 0 even with the best play:
https://lichess.org/analysis/standard/6k1%2F7p%2F1p4p1%2Fp1pp4%2F4r3%2F2PKP1P1%2FPP4P1%2F5R2_b_-_-_1_30
Thanks! I will remove that from the dataset.
@AnlamK said [^](/forum/redirect/post/Bu2Jw42u)
> I don't know how to link to an individual puzzle but for this one, if you let Stockfish think long enough, the evaluation drops to 0 even with the best play:
>
> https://lichess.org/analysis/standard/6k1%2F7p%2F1p4p1%2Fp1pp4%2F4r3%2F2PKP1P1%2FPP4P1%2F5R2_b_-_-_1_30
Thanks! I will remove that from the dataset.
@neilgd72 said ^
I don't have time, but if anybody wants to use Claude and the MCP above to auto-generate the rationales for the moves then I would be happy to donate some Claude credits and/or compute time /cc @FinalSpot.
Just PM me. You'd need to open source them and let me add them to the repository though.
I have updated to pre-generate all analysis and with Stockfish integration. The quality should be much better now, though still worse than a strong master analysis.
@neilgd72 said [^](/forum/redirect/post/llhMo4bm)
> I don't have time, but if anybody wants to use Claude and the MCP above to auto-generate the rationales for the moves then I would be happy to donate some Claude credits and/or compute time /cc @FinalSpot.
>
> Just PM me. You'd need to open source them and let me add them to the repository though.
I have updated to pre-generate all analysis and with Stockfish integration. The quality should be much better now, though still worse than a strong master analysis.
Great idea and execution. Unfortunately, the puzzles are too difficult for me. I couldn't really understand why the right move was the right move in most cases. Maybe later in my chess journey. Or perhaps there are teaching resources hidden away in the Studies that teach the concepts necessary to solve the puzzles?
Great idea and execution. Unfortunately, the puzzles are too difficult for me. I couldn't really understand why the right move was the right move in most cases. Maybe later in my chess journey. Or perhaps there are teaching resources hidden away in the Studies that teach the concepts necessary to solve the puzzles?