I just played a game and the analysis doesn't record any blunders for my opponent - at one point the analysis goes from -11 to -19 and it's like, "This is fine." and then at another point I am gifted a whole knight free.

Here is the game, any explanation appreciated: lichess.org/I37AVoxq

1

Lichess switched from using raw centipawn loss when analysing mistakes to a formula that calculates a metric called “winning chances” because going from -11 to -19 when you're losing doesn't change your winning chances – you're still losing.

Formula: wChances = 2 / (1 + Math.exp(-0.004 * cp)) - 1

Lichess calculates current chances minus previous chances and if that difference is positive, it grades the result on a scale (Innacuracy is .1, Mistake is .2, Blunder is .3)

In your case, the delta is negative so the move is not evaluated as a mistake.

81

-11 to -19 is not recorded as a blunder because it is both completely winning for black. it's the way it is programmed

3

typical 1800 rapid game

37

Thanks @thaurelia - I understand now. I always thought it was the delta between evaluation shifts, so was really surprised when I checked the final analysis and it didn't indicate _any_ blunders.