Is there a relationship between ELO and average centipawn loss?

Hello chess.com members,

Recently I analysed one of my games on one of chess.com's rival websites, leechess.com. Their chess analysis is free and currently I cannot afford membership on chess.com anymore. It is also pretty much instantaneous so, you don't have to wait unlike on chess.com.

But the point of this post is not to bash on chess.com's computer analysis feature. Rather, it is my goal to suggest to you all a little research experiment that I would like to perform, yet I have no idea how to accomplish it. I have the idea, but I'm not sure how to carry it out entirely. That's where you wonderful people come in!

On leechess.com, I saw an feature called Average Centipawn Loss and after further research I have a very vague understanding of what it means - namely, that it represents the average loss of 1/100 of a pawn in advantage (therefore centipawn) for each move one plays.

This led me to think about whether or not there may be a strong relationship between ELO and ACL (average centipawn loss abrieviated to this from here on out), since as one becomes stronger, one's ELO rating increases and the ACL drops, because one plays more accurately. But upon researching, I found that no one has ever attempted to plot ELO and ACL against each other.

This led me to ponder about how one could attempt to plot the two against each other. Chess.com, and many other websites, have humongous databases full of games with the respective ELOs of the players involved. If each of the games were analysed by a chess engine, either locally on stockfish or whatever software chess.com uses, you could get the ACL for each player. This data could then be plotted in Excel and you could then see what sort of relationship the two form. However, I don't know a fast way to access all of the computer analysis of all the games in the database. If someone has any idea how to accomplish the data getting/grabbing, then the Excel part is almost trivial.

Any ideas? I think this could be a potentially very interesting method of looking at rating and all help is welcome.

http://chess-db.com/public/research/qualityofplay.html

Here come ur second extended essay :) 

why not first, simply draw a scatter graph between ELO and ACL, and see the trend. Any great mathmathicien starts from this to further devolop their thinking 

IF there is no relationship at all, ur thinking is wrong but if u can see slight trend, then we can further think about it 

Yea i know how to make a scatter graph. What I care about is getting the data.

As micberu points out, this has been done already.

You'll get quite a lot of results if you google this - they use this method to try to compare players from different eras.

The problem has always been that it's much easier to play the top move or nearly the top move in simple positions than in complicated ones.

This is a good one, suggests that Fischer was stronger than Kasparov:

http://web.tecnico.ulisboa.pt/diogo.ferreira/papers/ferreira12strength.pdf

these studies never talk about different time controls for example the above link http://chess-db.com/public/research/qualityofplay.html talks about the correlation between quality of play and elo but they don't say at what time control they were measuring these games. The q.o.p. of  any given player is going to be different in blitz or bullet than it would be in a 30 minute rapid game. I don't see any mention of this though. I would like to know what the avg centi-pawn loss for say a 1200 player is in 1 min, 5 min, 30 min. etc.

There is a relationship between average centipawn loss and polygamy.   You play to win with extra pawns, then leave enemy only a king while you promote those pawns to queens to gain a very high score.

There is also a relationship between average centipawn loss and resigning.  Evil players who hate polygamy resign before your king can have the bliss of having many queens/wives.  This can also be exploited by a newbie, play a few good starting moves that lead to best gain in rating possible with the computer chess engine, then resign to have a centipawn score that rivals the best grandmasters.

Yes, there can also get a relationship between elo and average centipawn, but only if not already in a committed relationship with polygamy, resignamy, etc.