Відео

Thanks

To see how the depth increases playing strength

I certainly could, but what is poor-tuned and fine-tuned settings according to you?

I could have, but channels that provide accuracies after 2 opponents with a rating similar to the analyzing chess engine often give misleading results. What I mean is, accuracy metrics are only meaningful when analyzing games between two opponents weaker than the analyzing engine itself. In this case, Stockfish 17 at depths 20 and 60 are both stronger than Stockfish 16 at the depth used for analysis. This makes the accuracy and comparison of the playing strength inaccurate because the analysis engine doesn’t fully capture the deep nuances that Stockfish 17 does. The accuracies of their play according to the weaker engine was at 96-97% on both sides fyi. It doesn't actually mean anything though, because they both played stronger than the analyzing engine.. Also, the accuracy of the same game analyzed by different chess engines will always give different results... We could only really assign accuracy numbers that have a meaning if we had solved chess entirely, meaning we could actually compare it to the most perfect play

Yeah I use that site for free game reports

The version that evaluates the game is Stockfish 16 at depth 20. chess.wintrcat.uk/

lmao

Possibly

@@Stockfisherprobably

wait I recognize this board & eval. This is the free game review website called game report

I totally believe you

what is that?

@@TheBcoolGuy queens gambit declined. the structure where white plans d4 and c4 and black goes d5 and e6. It's also possible for black to go c6 (Slav) or both (semi-slav).

Didn't see mate in 61

diminishing returns beyond depth 20 for most engines.

Both sides are running the same Stockfish 17 engine, meaning they use identical algorithms to evaluate positions and search for moves. While Depth 60 searches deeper and can refine its understanding of future positions better than Depth 20, the difference is often incremental rather than decisive. Depth 20, already seeing 10 full moves ahead, is incredibly strong and capable of finding highly accurate moves. Depth 60, on the other hand, can spot subtler improvements or anticipate deeper counterplay, which is why its evaluation from the opening to the middlegame improved from +0.14 to +0.50. However, this slight advantage was not enough to convert into a win, largely because both engines play extremely cautiously without an opening book, minimizing risks and often steering towards draws. In more imbalanced, complex, or dynamic positions, Depth 60 might have been able to outmaneuver Depth 20.

@@Stockfisher exactly, its quite literally comparing apples to apples, depth 20 is enough to get the gist of what the main line should be without searching all the way through it.

@@michn5711 Pretty much

Please recommend a completely free chess database which is better than the lichess database

@Stockfisher

@@masterofskills76 According to Chatgpt: SCID: Open-source, large game database, advanced analysis. ChessDB: Free, open-source, with engine support. OpeningTree: Open-source, useful for studying openings. FEN Viewer (ChessBase): Free access to a large collection of games.

Yeah saw that too, absolutely no reason for him to give his rook at 3:30

The bot was trained and made to play like Tal did when he was alive and come up with these tricky tactics and traps, which usually isn't the best move, but shows his creativity and competence at least. That alone is impressive honestly... But I just like that it shows us how good he was in his days, considering he couldn't study Chess like we can now.. For me the 'art' is that we're able to create and train a neural network to play Chess in a similar way the real person would, even after he's gone ig

I actually missed the queen sacrifice playing a 2+1 game, that's why the first two moves weren't the top engine move... I reviewed my game and then saw I could've sacced my queen, leading to a forced checkmate.. I still won the game because my position was great though

I made this video because a viewer specifically requested it in my comment section... Obviously these extremely low elo bots will make huge mistakes repeatedly...

It was painful to watch, especially the endgame haha

What do you mean?

Stockfish depth 1 is actually pretty good for being only able to see 1 full move ahead, but will blunder mate in 2 and above because it can only see 2 plies ahead (one move by each side). People estimate its elo to be around 850-1000.

I don't think depth 1 is 1000 elo. It is 1600-1700 approx because depth 5 is around 2300

@@TastyMacronii Well, SF depth 1 strength varies massively patch to patch because it is not meant for this.. It also highly depends on the current position. SF at depth 1 will easily miss reasonably easy to spot tactics 1600-1700 would find most of the time due to it's inability to see further than 2 full moves ahead

Alright, I'll make more content in the future with imbalanced openings but have a really busy schedule for the next 3 weeks... Do you have any bots specifically in mind?

@Stockfisher Hikaru, Anand, Judit, Pia, Danya.

And Rey Enigma

I really hope you get well results from your exams.

@@Nottheworst4565 Appreciate it