Game complexity metric

@CaseyReese That is awesome! I do data science for a living and this is such a perfect use of this technique. Really appreciate the post.

You have asked about game complexity. But a slightly different question has interested me. I am interested in position complexity because I would like to use such a metric for automatically discovering complex puzzles (as opposed to puzzles that use number of moves to mate as a metric). I have read somewhere that stockfish (and I am sure many other engines) use some metric to decide if a position needs more analysis. They may be using how much the evaluation fluctuates during the calculation. I believe that it is possible to print out these numbers by interacting with stockfish from outside (i.e., without having to enter the code). I have not tried it. But some scripts by Ferdinand Mosca https://github.com/fsmosca may be using it. I would take a look at the script chess-artist on his github page.

Another thought I have about complexity is that if analysing a position requires looking at more width of the search tree as opposed to the depth (say, for the same number of nodes visited), then I find the position complex. For example, there are 'mate in 2' puzzles (usually composed, e.g., some problems by Sam Loyd come to mind) that can be extremely difficult. I think this can be programmed within an engine.

Is weighted error value supposed to help with this?

There is no strong metrics for measuring positional complexity because chess engines are not programmed to do anything like this. However, Maia's blunder rate calculation (look at their paper) is a good proxy.

YOU make it complicated when it can be less complicated by your own choice by being "macho" lol . Chess has evolved towards new complexities of material & positional play but so many still wanna just trap & "make" everything more dynamic whenever there is a clear solution ... sure people make more mistakes in 'complex' positions go ahead & have a good time' Just please don't imply Kasparov who had to DEFEND 20+ Draws when he had his back to the wall & Alekhine who had to play endings as well as Capablanca in their Match or Fischer Carlsen with their "simple" style at time are any less than say Mikhail Tal who by the way could play ALL Positions well also . Feels like a waste of my breath at times convincing people Leave the rules of Chess Alone . I think I went off subject a bit' but heh' I'm human in a world dealing with alot hehe !

@ThunderClap In this thread specifically, this would be considered "off-topic" because the subject matter involves a ratio of centipawn loss to game complexity (and not just game complexity as a primary measure)

To resolve your matter, I think it'd be possible to take into consideration a point wherein game complexity seems to have a lesser effect on blunder rate using ELO. I believe the point you're trying to make is that complexity at the GM level doesn't seem to have much of an effect on CPL, therefore some curve relating to the above should precipitate as a function of increasing ELO.

For the vast majority of chess players, particularly those who are looking to measure their progression outside of ELO, it seems many would want to determine their CPL as a function of complexity. This also measures absolute progress on "simple" games as well. We could even call it a function of simplicity if that'd make you feel better XD

If a person TRIES to make Chess even more complicated than it is (When there are other ways) then if they make mistakes the centipawn loss was of their own RESPONSIBILITY

@ThunderClap I'm not sure what you're on about here, but it really feels like you're referencing things outside of chess here..... while I empathize, please take venting elsewhere :)

Bumping this thread for discussion.