Very low centipawn loss?

If you check this you have my acpl by piece moved.

lichess.org/insights/fonsecasf/acpl/piece/phase:1/variant:rapid

You have: 26,3 for knight and 43,9 for bishop. Clearly i am playing with more precision my knights than my bishops. Rook have 50,1 and queen 71,2. I may say that i don't play as well with the heavy weights compared with lower pieces. Best value for king with 18.8. I may say i understand and move the king at the right times and with precision. I am not questioning if 18,8 is better than yours 24,9 for example, because the way we play the game may be very different. But may give hints to improve myself.

@fonsecasf We can't look at your "chess insights".

@A_0123456 didn't knew that... but i wrote down the values, so i hope you understood my point

@Morozov ""flaws of Lichess computer analyses" are next to nonexistant"

Have you ever tried to compare the values given by server analysis with the ones you get when you let your own machine separately calculate each move for even only 10 seconds, let alone longer?

Recent extreme example where server analysis even misses a mate in one (!!?):
lichess.org/forum/lichess-feedback/faulty-computer-analysis

Rare case, but still. This behaviour can certainly be called "flaw".

About ACPL:

If you average that average over many, many games, it's certainly possible to see some trends. But as for an individual game, @Chuck_Fess is right - you cannot really derive much from the number alone (even if server analysis was more precise). Except the obvious "the winner has a lower one than the loser". Or if it's really high (you don't get 100+ ACPL without some mistakes). But an ACPL of 20 could be anything between a one-sided massacre and a dull draw.

@ProfDrHack "But an ACPL of 20 could be anything between a one-sided massacre and a dull draw."

that's true. it is more interesting in the long run, because individual values tends to be inexpressive. So a trend may come up. Did not know that lichess analisys even misses mate in one though...

"Have you ever tried to compare the values given by server analysis with the ones you get when you let your own machine separately calculate each move for even only 10 seconds, let alone longer? "
Yes my point still stands, server mistakes are next to nonexistant. This is a bug but in the end server analyses tends to be far deeper than my computers (i5 and i7) can do in a reasonable amount of time.

Your comment about ACPL shows that you have not read the previous posts properly. That is exactly what I argued: If I can properly classify games by their different qualities and compare apples to apples and oranges to oranges it is possible to discern patterns and subsequently draw conclusions. And even for a single game it is entirely possible to use that number E.G as a confirmation for your intuition, if you have the necessary expierence with lichess ACPL.