We are in a time when computers win to chess masters giving knights odds and beginners get master titles in correspondence chess championships sending away the computer moves. Even so you write such a prose because you feel inspired today. Ok. That's enough for me.
Is there a software that can rate your games?
To finish, I'll say this: if you're asking for softwares that do give some accuracy % to chess games, like that gadget on chess.com does, that's a different problem. Ratings are ratings, and % of accuracy (like how close to the best move the software would play) is an other thing.
Agreed.
You can, of course, do a mathematical least-squares regression on the table of "typical ratings vs typical accuracy%" statistics and thereby derive an equation to convert your current game's accuracy% into something you could call a "Rating"... but the whole idea is silly.
Statistics don't APPLY to individual cases, such as a randomly-chosen chess game.
An 1800 player might play one game at 2150 strength and then play the very next game at 1450 strength. By looking at a single game, you derive a "rating" that means nothing. By looking at TEN games, you derive a "rating" that means next-to-nothing. You would need a much larger data set before a statistical approach can yield useful results.
To finish, I'll say this: if you're asking for softwares that do give some accuracy % to chess games, like that gadget on chess.com does, that's a different problem. Ratings are ratings, and % of accuracy (like how close to the best move the software would play) is an other thing.
Agreed.
You can, of course, do a mathematical least-squares regression on the table of "typical ratings vs typical accuracy%" statistics and thereby derive an equation to convert your current game's accuracy% into something you could call a "Rating"... but the whole idea is silly.
Statistics don't APPLY to individual cases, such as a randomly-chosen chess game.
An 1800 player might play one game at 2150 strength and then play the very next game at 1450 strength. By looking at a single game, you derive a "rating" that means nothing. By looking at TEN games, you derive a "rating" that means next-to-nothing. You would need a much larger data set before a statistical approach can yield useful results.
Analyzing one game the computer gives you a % of accuracy for that game and an Elo estimation for your play in that game, and not for the games he didn't analyse, one don't even need to be "clever" to understand this.
And you need to be clever to understand the software has no idea of what an ELO rating is, and it seems to be the same for you.
Elo ratings is maths based on data. The only data that is reliable and used for the ELO rating, is the chess RESULTS, not the quality of play nor the accuracy of moves. You win with bad moves? You gain ELO rating points. You lose after playing super good moves? You lose ELO rating points.
Here a link to the Wikipedia page in Portuguese:
https://pt.wikipedia.org/wiki/Rating_Elo
Maybe after reading this, you will understand the link you imagine between ELO ratings and "accuracy compared to computer play" is faint, if there is any.
I learned what an Elo is even before you got a mustache.
I never wrote about Elo calculation, I wrote about Elo estimation. Most of the people here can get the difference, even people with a fake mustache.
If all the OP wants is a way to convert accuracy% into a number that he can call a "Rating", then just run a least squares analysis on this sort of data table:
He would want to use data spanning a much wider range of ratings, though.
If all the OP wants is a way to convert accuracy% into a number that he can call a "Rating", then just run a least squares analysis on this sort of data table:
He would want to use data spanning a much wider range of ratings, though.
This is a very good work in my view. Alekhine games were brilliant but his accuracy was a bit lower than Karpov so he gets a lower rating in Elo estimation.
I posted it because I thought you might find it interesting.
Whether it's actually USEFUL or not is a seperate question. An equation derived from curve-fitting sixteen data points is little better than a wild guess. Especially since all sixteen of the data points fall within a four-point CAPS accuracy span.
Allowing for the inherent inaccuracy in extracting a CAPS accuracy score, there are a hundred different curves you could draw through those data points.
Sure MelvinGarvey, anyway the question was about software that can rate play in a single game, not the accuracy of that rating; for accuracy you need a good sample and easy statistics.
Erm, I don't know what means "can rate" in your mind, but to me, before we specify otherwise, it means "can rate efficiently enough". Or throwing a stone at a rock could be named backing.
Nowadays, computer strength and accuracy in chess is more that enough to rate a player in a game in percentage or Elo estimation. They are very close to perfection and you have to abide by their judgement.