Brilliant Moves in New Game Analysis Report

https://www.chess.com/analysis/game/computer/56648367?tab=review

11.. ....Bxh2+!!

https://www.chess.com/analysis/game/computer/56592771?tab=review

https://www.chess.com/analysis/game/computer/56652515?tab=review 7.....Bxa3 should be a brilliant

A brilliant move can actually be assessed by the engine. This topic is up to debate, but I had this proven when I faced off on a bot with assisted mode on. The engine's final decision was recognized as a brilliant move immediately. Therefore, I find that a brilliant move is something that looks a lot like a blunder, but the engine, looking ahead and seeing the advantage, shows that it is brilliant from the hidden advantage again. Usually, a brilliant move is a move that sacrifices a piece for more material, development, or checkmate much later on, typically without a clear path for a human to track, hence the "hard to find part", which usually happens to me within about 6 moves whenever I find one, such as a bishop sacrifice on the castled king.

https://www.chess.com/analysis/game/computer/56655269?tab=review

I don't see how...White simply blundered. Just because it's an obvious blunder doesn't mean your natural capture (not even a sacrifice) should be considered "brilliant".

No, brilliant move occurs when Chess.com cloud engine has a "sacrifice detection" on, which means they award piece sacrifices a lot (not including pawn sacrifice) . Even other tatics like zugzwang are not awarded chess.com "brilliancy". Furthermore, the brilliant move need not be the only winning move, and what more it may be the 2nd best move.

Also, when you run game review differently the results are differnet. For example, the app may show different move as brilliant compared to the browser. Furthormore, app self-analysis mode also has the feedbacks on, and they classify their move instantly, but can "change their mind". ALso on chess.com/event, chess.com runs Stockfish 15 NNUE that is in sync with the Game review. It is also more accurate when you run a higher depth of game review.

https://www.chess.com/events/2023-reykjavik-open/02/Williams_Simon_K-Rakhmangulova_AnastasiyaPressing on the "review" tab you see 2 brilliant move, which doesnt include leaving the queen en prise game review run by chess.com in sync with Stockfish 15 NNUE, whereas https://www.chess.com/news/view/2023-reykjavik-open-williams-sacrifices-queenthe article show someone personally game reviewing it having 3 brilliant move, including leaving the queen en prise, as leaving the queen en prise is not the top move recommended by the engine

Also, move classification on chess.com/events review tab and browser game review can NOT "change their minds" whereas the assistant against bots and app game review CAN "change their minds"

https://www.chess.com/analysis/game/computer/57023487?tab=review

I cant understand why 14. Rb8 and 15. b5 are brilliant

I think I have a basic idea of what the 'problem' is.

Let me address in two parts....just woke up so hopefully I can string together the thoughts reasonably. Oh, and I see this in the Chessbase 'Tactical Analysis' feature as well. In fact, I downloaded this game, ran it there and got the same kind of results...using the current developmental Stockfish and the most recent Komodo/Dragon engine.

1. The logic behind any engine you chose simply goes a bit 'haywire' in games with VERY weak players. The reason is the point evaluation for the 'better' side becomes SO HIGH so often because of the bad play, that the better side is 'spoilt for choice'. At that point you will almost always see MANY really good choices, while 1 option might be clearly better than the others...numerically, but not proportionately.

That is to say... a 'sliding scale' might should be implemented. I mean...lets say 5 moves give the side to play a 5.0 to 5.50 edge while a 6th gives a 6.5 edge that 1.0 difference is far less an actual difference than the difference between 5 moves giving the moving side a .25 to .33 edge while a 6th gives a 1.33 edge. This latter 1.0 difference actually matters a lot more towards the possible results of winning the game. And lets remember, these are very quickly analyzed....

2. The programming behind this particular analysis feature tilts toward 'piece activity'....so (coupled with the above) it finds those 'active' ideas even more desirable as they naturally tend to push the eval even higher compared to alternatives...which seems bound up with the programming choices of output like what constitutes a '!!' at the time the engine moves on to another move in the game to analyze.

All this 'chess AI' is in its infancy. To rely so much on certain parameters based on numeric evaluation - across all ranges of chess levels of play - is going to give you 'weirdness'...especially among these VERY weakly played games.

No, nothing 'clever' about it. White simply played a bad move that lost a piece. See my post above.

So...why are you posting this?

i suppose u know better

this is showing you how to counter that philodor variation