Your network blocks the Lichess assets!

lichess.org
Donate

Science of Chess: Let your Brilliance be your guide

@tpr said ^

"games which were analyzed by the computer"

  • And how to know if it is the winner or the loser who requested the computer analysis?

This I have to look into because I don't use the Lichess API very much, but it could be that the record you get includes whether a user requested the analysis. Someone who knows more, please feel free to weigh in!

@tpr said [^](/forum/redirect/post/EOOjGc6g) > "games which were analyzed by the computer" > * And how to know if it is the winner or the loser who requested the computer analysis? This I have to look into because I don't use the Lichess API very much, but it could be that the record you get includes whether a user requested the analysis. Someone who knows more, please feel free to weigh in!

"whether a user requested the analysis"

  • There are 8 possibilities:
    The winner requested the analysis and the loser looked at it as well.
    The winner requested the analysis and the loser did not look at it.
    The loser requested the analysis and the winner looked at it as well.
    The loser requested the analysis and the winner did not look at it.
    A kibitzer requested the analysis and both the loser and the winner looked at it as well.
    A kibitzer requested the analysis and the loser looked at it as well, but the winner did not.
    A kibitzer requested the analysis and the winner looked at it as well, but the loser did not.
    A kibitzer requested the analysis and neither the winner nor the loser looked at it.
"whether a user requested the analysis" * There are 8 possibilities: The winner requested the analysis and the loser looked at it as well. The winner requested the analysis and the loser did not look at it. The loser requested the analysis and the winner looked at it as well. The loser requested the analysis and the winner did not look at it. A kibitzer requested the analysis and both the loser and the winner looked at it as well. A kibitzer requested the analysis and the loser looked at it as well, but the winner did not. A kibitzer requested the analysis and the winner looked at it as well, but the loser did not. A kibitzer requested the analysis and neither the winner nor the loser looked at it.

Problem with studying chess is breaking it down into individual measurable components. As opposed to just a record of performance, we would need individual statistics separate from win loss and rating. Chess improvement also comes mostly from individual components, deliberate practice, not play, and practice is mostly separating skills into discrete units and working on them one at a time.

Although chess was interesting to psychological research due to the rating systems and multiple distinct levels where the superior player can win almost every game (each 400 rating points superior player wins over 90% of time with from beginner to Master at least 4-5 levels of the superior player winning over 90%), once other sports, music, math, professional expertise was studied, chess was less informative. That is largely due to the difficulty of measuring anything other that win / loss and rating.

Further study into chess and science of expertise, would need to focus on the individual components, how to measure the skill of the individual components, and how those individual components fit into rating.

As chess skill itself has failed to demonstrate the ability to transfer to other skills, it is possible that individual skills within chess might be developed within chess and transfer to other fields. If that is the case, would change chess in education, where getting better at chess overall has no practical benefit other than being good at chess, other useful skills can be developed by studying chess and transferred to other fields.

As I point out constantly to other chess trainers in education, the benefits are from studying chess, not playing chess. And in chess in education, chess is a bait and switch, to get children interested in developing study habits, specifically in the formative years of elementary and middle school, and there is almost no practical benefit from playing chess, other than playing chess can give incentive to study chess, and the study of chess can help one improve in certain mental skills that can transfer to other fields.

And the one last point, as I make often 'if you are normal and well adjusted, you don't need chess', as you can develop any of the skills from chess within a normal education. So adults trying to improve at chess, usually show a failing of learning and developing skills that were best developed as a child, and unless the adult is 'not normal and well adjusted', they should have already developed and worked on those skills as a child.

Problem with studying chess is breaking it down into individual measurable components. As opposed to just a record of performance, we would need individual statistics separate from win loss and rating. Chess improvement also comes mostly from individual components, deliberate practice, not play, and practice is mostly separating skills into discrete units and working on them one at a time. Although chess was interesting to psychological research due to the rating systems and multiple distinct levels where the superior player can win almost every game (each 400 rating points superior player wins over 90% of time with from beginner to Master at least 4-5 levels of the superior player winning over 90%), once other sports, music, math, professional expertise was studied, chess was less informative. That is largely due to the difficulty of measuring anything other that win / loss and rating. Further study into chess and science of expertise, would need to focus on the individual components, how to measure the skill of the individual components, and how those individual components fit into rating. As chess skill itself has failed to demonstrate the ability to transfer to other skills, it is possible that individual skills within chess might be developed within chess and transfer to other fields. If that is the case, would change chess in education, where getting better at chess overall has no practical benefit other than being good at chess, other useful skills can be developed by studying chess and transferred to other fields. As I point out constantly to other chess trainers in education, the benefits are from studying chess, not playing chess. And in chess in education, chess is a bait and switch, to get children interested in developing study habits, specifically in the formative years of elementary and middle school, and there is almost no practical benefit from playing chess, other than playing chess can give incentive to study chess, and the study of chess can help one improve in certain mental skills that can transfer to other fields. And the one last point, as I make often 'if you are normal and well adjusted, you don't need chess', as you can develop any of the skills from chess within a normal education. So adults trying to improve at chess, usually show a failing of learning and developing skills that were best developed as a child, and unless the adult is 'not normal and well adjusted', they should have already developed and worked on those skills as a child.

@tpr said ^

"whether a user requested the analysis"

  • There are 8 possibilities:
    The winner requested the analysis and the loser looked at it as well.
    The winner requested the analysis and the loser did not look at it.
    The loser requested the analysis and the winner looked at it as well.
    The loser requested the analysis and the winner did not look at it.
    A kibitzer requested the analysis and both the loser and the winner looked at it as well.
    A kibitzer requested the analysis and the loser looked at it as well, but the winner did not.
    A kibitzer requested the analysis and the winner looked at it as well, but the loser did not.
    A kibitzer requested the analysis and neither the winner nor the loser looked at it.

Unless the user who made the request is coded in the API data, which I don't know yet.

@tpr said [^](/forum/redirect/post/6ycMHAEW) > "whether a user requested the analysis" > * There are 8 possibilities: > The winner requested the analysis and the loser looked at it as well. > The winner requested the analysis and the loser did not look at it. > The loser requested the analysis and the winner looked at it as well. > The loser requested the analysis and the winner did not look at it. > A kibitzer requested the analysis and both the loser and the winner looked at it as well. > A kibitzer requested the analysis and the loser looked at it as well, but the winner did not. > A kibitzer requested the analysis and the winner looked at it as well, but the loser did not. > A kibitzer requested the analysis and neither the winner nor the loser looked at it. Unless the user who made the request is coded in the API data, which I don't know yet.

"Chess improvement also comes mostly from individual components, deliberate practice, not play"

  • Improvement comes from play, analysis of lost games, analysis of grandmaster games, and endgame study.
    Play without study is futile. Study without play is sterile.

"chess was less informative"

  • Chess is more informative: chess rating is more objective than math or music skill.

"individual skills within chess"

  • 'Education in Chess has to be an education in independent thinking and judging.' - Lasker

"the benefits are from studying chess, not playing chess"

  • Benefits of playing: making accurate decisions under pressure.
"Chess improvement also comes mostly from individual components, deliberate practice, not play" * Improvement comes from play, analysis of lost games, analysis of grandmaster games, and endgame study. Play without study is futile. Study without play is sterile. "chess was less informative" * Chess is more informative: chess rating is more objective than math or music skill. "individual skills within chess" * 'Education in Chess has to be an education in independent thinking and judging.' - Lasker "the benefits are from studying chess, not playing chess" * Benefits of playing: making accurate decisions under pressure.

@tpr said ^

"whether a user requested the analysis"

  • There are 8 possibilities:
    The winner requested the analysis and the loser looked at it as well.
    The winner requested the analysis and the loser did not look at it.
    The loser requested the analysis and the winner looked at it as well.
    The loser requested the analysis and the winner did not look at it.
    A kibitzer requested the analysis and both the loser and the winner looked at it as well.
    A kibitzer requested the analysis and the loser looked at it as well, but the winner did not.
    A kibitzer requested the analysis and the winner looked at it as well, but the loser did not.
    A kibitzer requested the analysis and neither the winner nor the loser looked at it.

OK, edited the blog post to confirm (and discuss) that it is indeed the case that the data does not have info about who requested the analysis!

You can see in the revised post that I talk about this in this fashion, but the gist of the author's point is that you still have to try and explain why you would see the positive effect of "win" analyses being requested on the winning player's ELO. That is, they are suggesting there is a "signal" in their data related to the winning player requesting an analysis of their own game amidst the "noise" of the other ways that analysis might have been requested.

@tpr said [^](/forum/redirect/post/6ycMHAEW) > "whether a user requested the analysis" > * There are 8 possibilities: > The winner requested the analysis and the loser looked at it as well. > The winner requested the analysis and the loser did not look at it. > The loser requested the analysis and the winner looked at it as well. > The loser requested the analysis and the winner did not look at it. > A kibitzer requested the analysis and both the loser and the winner looked at it as well. > A kibitzer requested the analysis and the loser looked at it as well, but the winner did not. > A kibitzer requested the analysis and the winner looked at it as well, but the loser did not. > A kibitzer requested the analysis and neither the winner nor the loser looked at it. OK, edited the blog post to confirm (and discuss) that it is indeed the case that the data does not have info about who requested the analysis! You can see in the revised post that I talk about this in this fashion, but the gist of the author's point is that you still have to try and explain why you would see the positive effect of "win" analyses being requested on the winning player's ELO. That is, they are suggesting there is a "signal" in their data related to the winning player requesting an analysis of their own game amidst the "noise" of the other ways that analysis might have been requested.

"there is a "signal" in their data"

  • Maybe repeat for 15+10 time control and no rating limit to see if a similar signal occurs.
"there is a "signal" in their data" * Maybe repeat for 15+10 time control and no rating limit to see if a similar signal occurs.

@tpr said ^

"there is a "signal" in their data"

  • Maybe repeat for 15+10 time control and no rating limit to see if a similar signal occurs.

Sure. I'd even be happy to see if they can replicate this result by choosing different months, or considering a wider time interval that doesn't include this May 2023 data.

@tpr said [^](/forum/redirect/post/x0hygx2y) > "there is a "signal" in their data" > * Maybe repeat for 15+10 time control and no rating limit to see if a similar signal occurs. Sure. I'd even be happy to see if they can replicate this result by choosing different months, or considering a wider time interval that doesn't include this May 2023 data.

@NDpatzer

they are suggesting there is a "signal" in their data related to the winning player requesting an analysis of their own game amidst the "noise" of the other ways that analysis might have been requested.

Maybe Gaining Elo may make you more motivated to see how accurately you played. Players who lose Elo overall may be less likely to click the computer analysis button after winning due to being less motivated. Players who lose Elo may go on tilt and may quickly play more games to try to gain rating without bothering to click the analysis. I observe this in myself.

If this occurs then it would cause a greater association between greater increase in Elo and clicking the computer analysis button after wins. This means that computer analysis after winning may not cause more Elo increase, but more Elo increase leads to more computer analysis after winning.

That's a thought. Also a lot of players on Lichess don't do active practice. I played a lot without studying. Only recently I started doing Puzzle Racer. So when I clicked the computer analysis button it was because I was just curious. Without actually bothering to study. Maybe gains in Elo motivate looking at computer analysis after winning.

@NDpatzer >they are suggesting there is a "signal" in their data related to the winning player requesting an analysis of their own game amidst the "noise" of the other ways that analysis might have been requested. Maybe Gaining Elo may make you more motivated to see how accurately you played. Players who lose Elo overall may be less likely to click the computer analysis button after winning due to being less motivated. Players who lose Elo may go on tilt and may quickly play more games to try to gain rating without bothering to click the analysis. I observe this in myself. If this occurs then it would cause a greater association between greater increase in Elo and clicking the computer analysis button after wins. This means that computer analysis after winning may not cause more Elo increase, but more Elo increase leads to more computer analysis after winning. That's a thought. Also a lot of players on Lichess don't do active practice. I played a lot without studying. Only recently I started doing Puzzle Racer. So when I clicked the computer analysis button it was because I was just curious. Without actually bothering to study. Maybe gains in Elo motivate looking at computer analysis after winning.

@NDpatzer

Thanks for addressing the analysis concern in the blog. I have some thoughts to share. I hope you'll share what you think.

There are probably some ways to tell a different story (and if you have one, I'd be curious to hear it in the comments)

The big idea here is that if failure really does teach us more than success, higher rates of analyzing losses should predict greater positive change in ELO over time.

I don't agree and I'll tell you why. Firstly 'teach' is equated with clicking the computer analysis button.

Secondly, correlation is not causation. Computer analysis after winning may not cause more Elo increase, but more Elo increase may lead to more computer analysis after winning.

Even if you are uneasy about this aspect of the methods, how do you explain the observed effects (and the observed null outcomes)?

Gaining Elo may make you more motivated to see how accurately you played. Players who lose Elo overall may be less likely to click the computer analysis button after winning due to being less motivated. Players who lose Elo may go on tilt and may quickly play more games to try to gain rating without bothering to click the analysis.

I observe this in myself.

I think that most people who click the analysis button do not seriously use it to study. Even I don't really use it to study, just to see the accuracy and get a ego boost.

You said there were two caveats: that Elo Gain is associated with analyzing wins with up to a point (for 24% wins analyzed) and that the author did not have access to who requested the analysis, only that an analysis was requested.

The biggest caveat is the fact that clicking the computer analysis button is being equated with 'studying'.

That's the bigger problem.

What do you think?

@NDpatzer Thanks for addressing the analysis concern in the blog. I have some thoughts to share. I hope you'll share what you think. >There are probably some ways to tell a different story (and if you have one, I'd be curious to hear it in the comments) --- >The big idea here is that if failure really does teach us more than success, higher rates of analyzing losses should predict greater positive change in ELO over time. I don't agree and I'll tell you why. Firstly 'teach' is equated with clicking the computer analysis button. Secondly, correlation is not causation. Computer analysis after winning may not cause more Elo increase, but more Elo increase may lead to more computer analysis after winning. --- >Even if you are uneasy about this aspect of the methods, how do you explain the observed effects (and the observed null outcomes)? Gaining Elo may make you more motivated to see how accurately you played. Players who lose Elo overall may be less likely to click the computer analysis button after winning due to being less motivated. Players who lose Elo may go on tilt and may quickly play more games to try to gain rating without bothering to click the analysis. I observe this in myself. I think that most people who click the analysis button do not seriously use it to study. Even I don't really use it to study, just to see the accuracy and get a ego boost. --- You said there were two caveats: that Elo Gain is associated with analyzing wins with up to a point (for 24% wins analyzed) and that the author did not have access to who requested the analysis, only that an analysis was requested. The biggest caveat is the fact that clicking the computer analysis button is being equated with 'studying'. That's the bigger problem. What do you think?