- Blind mode tutorial
lichess.org
Donate

Estimate ELO rating based on stockfish analysis?

#1

Is there a way that when you load a PGN you can get a computer analysis of what the estimated ELO of the black and white pieces are? Sometimes I feel like I'm really strong for a ~1200, but other times I will play a game and not blunder anything and its very hard to tell if my opponent is really good against me or I am really good against them and their stronger than what they're rated. I wonder what the computer would rate both pieces what their estimated ELO/skill level for that game was. If there is a way to do that manually or through the use of another site feedback would be appreciated, I think this could be a cool feature to add to the analysis board aswell imo.

Is there a way that when you load a PGN you can get a computer analysis of what the estimated ELO of the black and white pieces are? Sometimes I feel like I'm really strong for a ~1200, but other times I will play a game and not blunder anything and its very hard to tell if my opponent is really good against me or I am really good against them and their stronger than what they're rated. I wonder what the computer would rate both pieces what their estimated ELO/skill level for that game was. If there is a way to do that manually or through the use of another site feedback would be appreciated, I think this could be a cool feature to add to the analysis board aswell imo.
#2

The thing is ELO is based on playing very many games. Sometimes a person can play very well for a few games, but most of the time they don't play that well, or vice versa. On a good day I might hang with a 1900, on a bad day I might lose to a 1000. So that actually reflects my rating that is somewhere in the middle. If I go one day and beat a few 1700s I cannot claim that as my rating because I'm not consistent at it. I beat SF level 6 a few times, rated 2300. I also lost hundreds of times........the reality is is that I'll probably NEVER beat that level 50% of the time as I started playing chess at age 46 and am 50 now, and still have a "day job".

The thing is ELO is based on playing very many games. Sometimes a person can play very well for a few games, but most of the time they don't play that well, or vice versa. On a good day I might hang with a 1900, on a bad day I might lose to a 1000. So that actually reflects my rating that is somewhere in the middle. If I go one day and beat a few 1700s I cannot claim that as my rating because I'm not consistent at it. I beat SF level 6 a few times, rated 2300. I also lost hundreds of times........the reality is is that I'll probably NEVER beat that level 50% of the time as I started playing chess at age 46 and am 50 now, and still have a "day job".
#3

interesting idea. elo is just based on how often you win and lose against others. Its a statistical score. The question is how do you relate the moves in a particular game to a certain playing strength? Since elo is just relative, you'd need to compare the moves made in your game to a database of games from the same group of players whose elo you want to base the evaluation on. I wonder if an AI could learn this? You train it to predict the elo of the players based on the moves in the game, kind of similar to the 'guess the elo' meme currently in vogue with a number of chess streamers.

interesting idea. elo is just based on how often you win and lose against others. Its a statistical score. The question is how do you relate the moves in a particular game to a certain playing strength? Since elo is just relative, you'd need to compare the moves made in your game to a database of games from the same group of players whose elo you want to base the evaluation on. I wonder if an AI could learn this? You train it to predict the elo of the players based on the moves in the game, kind of similar to the 'guess the elo' meme currently in vogue with a number of chess streamers.
#5

@ChaBoiNeil said in #1:

Is there a way that when you load a PGN you can get a computer analysis of what the estimated ELO of the black and white pieces are?

Yeah, there is. Multiply the inaccuracies with 50, the mistakes with 100 and the blunders with 200. Multiply the ACL with 10. Add all the numbers and subtract the sum from 2855. Presto, there's your Elo per game.

(Yes, I just made this up.)

@ChaBoiNeil said in #1: > Is there a way that when you load a PGN you can get a computer analysis of what the estimated ELO of the black and white pieces are? Yeah, there is. Multiply the inaccuracies with 50, the mistakes with 100 and the blunders with 200. Multiply the ACL with 10. Add all the numbers and subtract the sum from 2855. Presto, there's your Elo per game. (Yes, I just made this up.)
#7
<Comment deleted by user>
#8

Chess skill is not measured by looking at one game. It's measured by looking at both your good and bad games.

Your rating is a perfect measure of your playing strength. No other measure will ever be better.

Two 2000 rated people will on average score 0.5 points per game.

The same prediction cannot be made if you compare two people with 41 acpl.

Chess skill is not measured by looking at one game. It's measured by looking at both your good and bad games. Your rating is a perfect measure of your playing strength. No other measure will ever be better. Two 2000 rated people will on average score 0.5 points per game. The same prediction cannot be made if you compare two people with 41 acpl.

This topic has been archived and can no longer be replied to.