Quality Rating System !

Known also as QRS rating. :) Is rating system that take in count, Not the win, but the average centipawns loss that will give the currently best chess engine. (The end of flagging)

I see a serious problem with this. First of all, it depends strongly on the openings you play. Secondly, it depends even more strongly on how many moves are played in a given game. So you will get a much lower ACP if you play for say 50+ moves. And then how would you deal with time controls? Just too many problematic variables here that make this poor as a rating system.

...for the time control: When the time of a player end the game is over and the result is estimated (according to their ACPLs)

...as for the dependance of the opening: It will be another rating mode [Casual, Glicko, Lichess (or Stockfish) ] so you can adjust your openings when you choose it.

...as for the number of moves: The ACPL estimation confidence will varies according to the number of the moves.

If the Coders (respect) achieve to code it, will be expansion in the dimension of chess rating.

#2 I agree that it wouldn't be a "rating" system for purposes of comparing players to each other.

#1 Replace the word "rating" with the word "metric" and I strongly agree -- players with additional metrics at their disposal can better measure their progress over time.

I propose the following metrics (which could be filtered the same ways that games are filtered):
* Opening average CP loss and IQR CP loss
* Middlegame average CP loss and IQR CP loss
* Endgame average CP loss and IQR CP loss

IQR = interquartile range (plus median): for example, in the set:
(0, 25, 50, 75, 100)
the IQR plus median is:
(25, 50, 75)

and in the set:
(0, 10, 20, 30, 100)
the IQR plus median is:
(5, 20, 65)

Average CP loss is mainly useful for identifying games in which a player blundered. IQR values are useful for identifying overall (ignoring blunder outliers) strength and consistency of play.

Reading Toadofsky (#7) I realise that the first step is IQR metrics (middle 50% of the Bell graph) of the moves. And in the future maybe live rating. Underline rating (evaluation) because in this case the player approach to the game will change! The player will concetrate his effort in finding the best moves without carrying about losing on time or losing point of his rating if he is mated.

When one of the opponents feel confident that have better IQR-CPL he can state ''score'' or continue playing for the audience.
In this case a lot of jerks will come out on the surface.
At the end of the tournament the winners are two: the top scorer (the materialist) and the gentleman (the humanist – the one that people watching online had voted).
This will bring chess playing behavior closer to real life.
The good guys will be not top rated but will be people's heroes. And then there is possibility the world championship candidates to be voted by the people.
The World Chess Championship will be the Battle between God and Evil over the board.

I don't think statistics like this help anyone that much, but either way I love statistics, every statistic available is a great thing, just to look at and ponder. That said there are a few problems, it would mean that the server would have to analyze every single game (which would be very taxing), otherwise it would only show average centipawn loss for games players decided to analyze themselves. One way around this, would be to only include (what lichess considers) "classical" time control games, and have all classical games automatically analyzed. It's a bit silly to do serious analysis on bullet anyway.

Analyzing all classical games automatically would have the added benefit of more quickly identifying cheaters. Anyway, it's a good idea, but I'm not sure how high up the priority list it should be, and how taxing it would be on the servers.