lichess.org

ELO VS CENTIPAWN LOSS. Make a ranking of player taking into account their centipawn loss

Centipawn loss is absolutely meaningless outside of the context of the individual game for which it is recorded. Comparing centipawn loss between two different games by the same person is pointless, comparing centipawn loss between two different games that have no players in common is even more so.

Let's say two games have an identical position with white in a commanding lead; one black player resigns, while the other plays on for another 40 moves until white delivers checkmate. The latter white player (both black and white, really) has had 40 moves of basically zero centipawn loss to lower their average, while the former will see their average stay a bit higher, because there were not 40 moves with basically no wrong answer to pad the analysis. This doesn't even get into situations where the opponent blunders very early on, and the winning player's ACL is basically zero because they have +10 or so from the opening. ACL means absolutely nothing.

To be honest, I'm not even sure what it tells you about a single game. What inference can anyone reliably make based on ACL? I'd like it if someone could spell it out and tell me how they know that, this sort of topic comes up often and I never understand why people care. One inference that is definitely NOT reliable is, "low centipawn loss equals good play". Good play will probably produce a low ACL, but a low ACL can also be achieved by someone who plays terribly, because context matters.

@Chuck_Fess how dare you steal my answer 😛 yes I may agree but I'm also unqualified to say for positive.

@Chuck_Fess Most common reason for low centipawn loss is simple game. Lots of forced tradet etc will make loss small. Quality of play index does address some of concers you had. Not countinm moves make outside +-200 cp solves the problem someone resignin lost game and other one blundering to the end. So it coudl be used to estimated from low number of games Elo of a persdon faster thatn what Elo algorith can do. But probably about as quickly as Glicko-2

My centipawn loss for Blitz is 67.8 and I'm a terrible player - my rating is only 1170.

Like it has already been mentioned by @Chuck_Fess, it is an utterly pointless metric that has no correlation to the strength of players. I'd potentially have a higher ACL than Magnus Carlsen if I only played against 4-year olds and he only played against best players in the world.

You can't post in the forums yet. Play some games!

Reconnecting