Sometimes in games like this I'm wondering what a mistake actually is. Black made no mistakes in that game but what is a mistake anyway? Isn't there always a better move?
So what do you guys think and how does the lichess game analysis determine mistakes?
From my experience how this works is that Lichess sends the PGN through an engine and the engine returns positional analysis values at each move. Those values will range from anywhere as tiny as +/-0.00 to anything as large as +/-99.99 I believe (i.e. you're in a forced mating sequence). Obviously if the value returned by Stockfish is less than +/- 1, there is little advantage to doing thorough analysis because the position is still completely playable. A titled player could benefit from such changes in advantage, but a titled player would also likely want to have a more thorough analysis to determine if those small changes in value are accurate or not.
Anything with a value change of +/- 1.00 (i.e. the material value of a pawn) or more is at the very least an inaccurate move. That is why when the position returns such a value it is annotated. Anything with a value of +/- 2.00 or more (although this value might scale up as the game progresses, I'm not sure) is considered a "mistake" in the annotation. And anything with a large shift in the analysis (maybe +/- 4.00 or more?) is considered a "blunder" for the purpose of annotation.
So you can have a game where you make 30 inaccuracies and lose, but never make a mistake or blunder. Or you can have a game here 0 inaccuracies or mistakes are made, but one blunder is so bad you lose the game outright from it (like blundering your queen or moving into a forced mating line). I hope this helps?
Ah, good find Toad. I was a little off on my assessment then. Looks like 0.5-1 is inaccuracy, >1-2.99 is mistake, and >3 is blunder, with some modification for evals once the position is already +/-10.
