Accuracy Percentages

I was looking at my insights and noticed there's a tab for accuracy, which shows a percentage of the move's accuracy compared to a computer.
I switched to Lichess from Chess.com and one of the only things I've been having trouble getting used to is the centipawn loss instead of the out of 100 accuracy.
Having seen the accuracy in Insights, I was wondering if there's a way to have that instead of centipawn loss on the analysis board. Anyone know if a) it's possible and b) how to do it if it is?

> Having seen the accuracy in Insights, I was wondering if there's a way to have that instead of centipawn loss on the analysis board. Anyone know if a) it's possible and b) how to do it if it is?

It is not and I don't think it will be. Lichess has been regularly confused by chesscom accuracy metric since the parameters for the calculation of the number and the parameters to mark a move as brilliant or great or good are not known to the public.

I don't have it right here but I believe the acpl value calculation should be disclosed and published somewhere, which makes it more reliable.
Personally I don't focus too much on the number in itself, rather I do focus more on the number of mistakes and blunders.
Acpl is better when low but there are just so many factors that contribute to its final value such as your opponent's performance in the game and especially in the opening.

I wholeheartedly agree that Lichess should immidietly add a CAP score program in the analysis bar. It is much simpler and more helpful than the centi-pawn loss system that is currently in place.

Centipawn loss is pretty straightforward: it's how much your eval goes down per move.

Say the engine's best move had an eval of +1.36
Say you played a move with an eval of only +1.13
Then the centipawn loss is the difference: 23.

Then average centipawn loss is just the average across all the moves of the game. If it's zero, then you played perfectly according to the engine.

ex. If your ACPL is 45, that means your eval went down by -0.45 per move (on average).

To me, "accuracy" is more confusing—clearly 100% means the best engine moves, but what's 50%? What's 10%?

I guess you could set the worst move as 0% and the best move as 100%, but that means if you're in a situation where "every move wins" and you pick a slightly less winning move (say +9.5 vs +9.8) then your accuracy would be really low, even 0%! (vs cpl of 30 bc it was really close to the best move)

But anyway people seem to like the % system and I saw it too on Lichess recently so hopefully they came up with a neat solution! (maybe some kinda inverse logit transform over ACPL?)