- Blind mode tutorial
lichess.org
Donate

Science of Chess: What does it mean to have a "chess personality?"

Thanks, that was a great read and a sound approach to a very interesting problem, here's a couple of thoughts:

  • I'm not so sure the number of factors in a psychometric tool is more a statistical than a theoretical problem (I think they go hand in hand): of course once you have developed a survey, an exploratory factor analysis can point to a specific number of factors (although that stuff with eigenvalues is about rules of thumb and more or less arbitrary eye-tests), but you could argue that what's really decisive is what happens before that, i.e., the formulation of the items that can more or less explicitly be associated with a theoretical understanding of the phenomenon you're measuring.
  • Considering opening choices as a measure of chess personality, I'm afraid, assumes a bit too much about the awareness and willingness to accomodate one's personality of players at low-intermediate levels, and maybe is not a great idea for GMs and other high level players, as their choices might be due to considerations about typical (or specific) preparations of opponents.
  • I think, in order to disentangle skill and style, the only option is, as you suggest, to separately analyse players of different skill levels. However, I'm not sure using masters as a sample is the best way to distinguish between patterns in chess personality, as I suspect that their honed skills might result in a lower variability in measurable style. In other words, masters will often play the best move or one of the very few best moves, regardless of their preferences or styles, while patzers like us often won't calculate the correct option and this will allow them to show higher variability).
  • I think another interesting approach in establishing and validating tools to assess chess personalities could be to investigate the relationships between candidate measures and other psychological variables (e.g., of personality tests). For instance, if opening choices were strongly associated with personality traits from the big 5, that would point to them being an interesting measure of chess styles, besides suggesting that general and chess personalities might be related.
    Thanks again
Thanks, that was a great read and a sound approach to a very interesting problem, here's a couple of thoughts: - I'm not so sure the number of factors in a psychometric tool is more a statistical than a theoretical problem (I think they go hand in hand): of course once you have developed a survey, an exploratory factor analysis can point to a specific number of factors (although that stuff with eigenvalues is about rules of thumb and more or less arbitrary eye-tests), but you could argue that what's really decisive is what happens before that, i.e., the formulation of the items that can more or less explicitly be associated with a theoretical understanding of the phenomenon you're measuring. - Considering opening choices as a measure of chess personality, I'm afraid, assumes a bit too much about the awareness and willingness to accomodate one's personality of players at low-intermediate levels, and maybe is not a great idea for GMs and other high level players, as their choices might be due to considerations about typical (or specific) preparations of opponents. - I think, in order to disentangle skill and style, the only option is, as you suggest, to separately analyse players of different skill levels. However, I'm not sure using masters as a sample is the best way to distinguish between patterns in chess personality, as I suspect that their honed skills might result in a lower variability in measurable style. In other words, masters will often play the best move or one of the very few best moves, regardless of their preferences or styles, while patzers like us often won't calculate the correct option and this will allow them to show higher variability). - I think another interesting approach in establishing and validating tools to assess chess personalities could be to investigate the relationships between candidate measures and other psychological variables (e.g., of personality tests). For instance, if opening choices were strongly associated with personality traits from the big 5, that would point to them being an interesting measure of chess styles, besides suggesting that general and chess personalities might be related. Thanks again

Thanks for reading and for all these great comments! I've put a few responses below.

@jeffsonadam said in #2:

Thanks, that was a great read and a sound approach to a very interesting problem, here's a couple of thoughts:

  • I'm not so sure the number of factors in a psychometric tool is more a statistical than a theoretical problem (I think they go hand in hand): of course once you have developed a survey, an exploratory factor analysis can point to a specific number of factors (although that stuff with eigenvalues is about rules of thumb and more or less arbitrary eye-tests), but you could argue that what's really decisive is what happens before that, i.e., the formulation of the items that can more or less explicitly be associated with a theoretical understanding of the phenomenon you're measuring.

Yes, absolutely a fair point. Depending on both the method and the properties of the data, a purely numerical approach to choosing dimensionality absolutely should be informed by theory. Of course theory (depending on the problem) may also not lead to a specific prediction about this either, so an iterative approach and a willingness to consider alternative solutions is important.

  • Considering opening choices as a measure of chess personality, I'm afraid, assumes a bit too much about the awareness and willingness to accomodate one's personality of players at low-intermediate levels, and maybe is not a great idea for GMs and other high level players, as their choices might be due to considerations about typical (or specific) preparations of opponents.

I see the point, but I guess I also think players at all levels have variations they find more comfortable than others. Even if you're a very strong player prepping for a specific opponent you probably have options to consider as far as your response to their repertoire and perhaps this means you choose responses that steer you towards your type of game as much as possible. The 2023 paper I cite in the references has a number of neat things to say about opening repertoire as a function of skill level that are relevant to your point too. But yes - opening choice is not a final answer for sure, but perhaps a useful contributor?

  • I think, in order to disentangle skill and style, the only option is, as you suggest, to separately analyse players of different skill levels. However, I'm not sure using masters as a sample is the best way to distinguish between patterns in chess personality, as I suspect that their honed skills might result in a lower variability in measurable style. In other words, masters will often play the best move or one of the very few best moves, regardless of their preferences or styles, while patzers like us often won't calculate the correct option and this will allow them to show higher variability).

Yes, I very much agree about a sort of ceiling effect for very strong players. It's interesting though that GMs have been historically described in terms of personality a great deal. I don't know the extent to which this is true in the modern era - the rise of SF may mean that optimality is king.

  • I think another interesting approach in establishing and validating tools to assess chess personalities could be to investigate the relationships between candidate measures and other psychological variables (e.g., of personality tests). For instance, if opening choices were strongly associated with personality traits from the big 5, that would point to them being an interesting measure of chess styles, besides suggesting that general and chess personalities might be related.

Very neat idea and I'd be curious to see what comes out of such an analysis. Thanks again for reading!

Thanks again

Thanks for reading and for all these great comments! I've put a few responses below. @jeffsonadam said in #2: > Thanks, that was a great read and a sound approach to a very interesting problem, here's a couple of thoughts: > - I'm not so sure the number of factors in a psychometric tool is more a statistical than a theoretical problem (I think they go hand in hand): of course once you have developed a survey, an exploratory factor analysis can point to a specific number of factors (although that stuff with eigenvalues is about rules of thumb and more or less arbitrary eye-tests), but you could argue that what's really decisive is what happens before that, i.e., the formulation of the items that can more or less explicitly be associated with a theoretical understanding of the phenomenon you're measuring. Yes, absolutely a fair point. Depending on both the method and the properties of the data, a purely numerical approach to choosing dimensionality absolutely should be informed by theory. Of course theory (depending on the problem) may also not lead to a specific prediction about this either, so an iterative approach and a willingness to consider alternative solutions is important. > - Considering opening choices as a measure of chess personality, I'm afraid, assumes a bit too much about the awareness and willingness to accomodate one's personality of players at low-intermediate levels, and maybe is not a great idea for GMs and other high level players, as their choices might be due to considerations about typical (or specific) preparations of opponents. I see the point, but I guess I also think players at all levels have variations they find more comfortable than others. Even if you're a very strong player prepping for a specific opponent you probably have options to consider as far as your response to their repertoire and perhaps this means you choose responses that steer you towards your type of game as much as possible. The 2023 paper I cite in the references has a number of neat things to say about opening repertoire as a function of skill level that are relevant to your point too. But yes - opening choice is not a final answer for sure, but perhaps a useful contributor? > - I think, in order to disentangle skill and style, the only option is, as you suggest, to separately analyse players of different skill levels. However, I'm not sure using masters as a sample is the best way to distinguish between patterns in chess personality, as I suspect that their honed skills might result in a lower variability in measurable style. In other words, masters will often play the best move or one of the very few best moves, regardless of their preferences or styles, while patzers like us often won't calculate the correct option and this will allow them to show higher variability). Yes, I very much agree about a sort of ceiling effect for very strong players. It's interesting though that GMs have been historically described in terms of personality a great deal. I don't know the extent to which this is true in the modern era - the rise of SF may mean that optimality is king. > - I think another interesting approach in establishing and validating tools to assess chess personalities could be to investigate the relationships between candidate measures and other psychological variables (e.g., of personality tests). For instance, if opening choices were strongly associated with personality traits from the big 5, that would point to them being an interesting measure of chess styles, besides suggesting that general and chess personalities might be related. Very neat idea and I'd be curious to see what comes out of such an analysis. Thanks again for reading! > Thanks again

"Are you a player who is willing to forego castling in favor of a strong attach?" - attach should be attack if you want to fix typo

Actually as far as forced surveys go I think the problem that can sometimes occur is that sometimes people don't answer honestly but answer to get the result they want. For example if you identified as Belle as a Disney princess and were asked whether you'd consider dating someone who is abusive and on the hairy side you might answer yes to get Belle. A funny story happened when I knew someone taking a blind taste test for cokes and was a big Coca-cola fan so they gave one that they thought was Coca-cola 10/10 and rest got a 0/10 and then ruthless mocked them when they gave 10/10 to a different brand.

Actually for openings I've had reverse happen where people try to match my "Chess Personality" to get openings and suggest what to play based on that. So potentially someone could get do a quiz to get a personality, choose an opening based on it and then later you base their personality on the opening they chose based on their personality test.

There's also the issue that many assumptions about personalities don't necessarily hold true. For example Tal was well known for his tactical wizardry but isn't as well known for his endgames that he often played really well. In the same way Kramnik who is well known for his positional play has played some tactical games really well. So as you get stronger the concept of universal players does get more prominent.

Now as far as my suggestion for how to get a chess personality I would suggest the following approaches.

  1. Sharpness of positions reached is probably a good indication of how tactical a position is. This can be done with Leela's WDL score with people taking more risks having a higher sharpness score.
  2. For technical skills I think endgame accuracy could be a good metric. For example if you use 7 pieces for cut off to endgames and compare it to tablebase I think that might be useful as a metric potentially then modified for rating.
  3. For things like "Wildcard" and "Mad Scientist" you could use opening choice and see how often they play lines that aren't frequently used. If by move 5 or 10 they're in completely offbeat lines that would give a higher score there. Players like Basman would score really high there.
  4. I'm unsure if it's easy to see but if possible seeing where players win the games. If it's in opening then they're theoretical and do opening prep. If it's in middlegame then it's middlegame/attacking skills and if it's in endgame then they're technical players.
  5. Initiative/Aggression I think can be measured by non-material engine evaluations. For example if engine says I'm -0.5 but I'm a pawn down then I'm +0.5 in initiative and my opponent is -0.5 on initiative but +1.0 in material so would get a materialistic trait instead. Players who regularly get a +score excluding material would tend to be attacking players while those with a negative score I think tend to be defensive players. I would imagine most common positions would be ones where one side is +0.3 and has a slight edge but it's drawn with best play. This would also capture sacrifices to some extent.

Hopefully that's useful for providing metrics

"Are you a player who is willing to forego castling in favor of a strong attach?" - attach should be attack if you want to fix typo Actually as far as forced surveys go I think the problem that can sometimes occur is that sometimes people don't answer honestly but answer to get the result they want. For example if you identified as Belle as a Disney princess and were asked whether you'd consider dating someone who is abusive and on the hairy side you might answer yes to get Belle. A funny story happened when I knew someone taking a blind taste test for cokes and was a big Coca-cola fan so they gave one that they thought was Coca-cola 10/10 and rest got a 0/10 and then ruthless mocked them when they gave 10/10 to a different brand. Actually for openings I've had reverse happen where people try to match my "Chess Personality" to get openings and suggest what to play based on that. So potentially someone could get do a quiz to get a personality, choose an opening based on it and then later you base their personality on the opening they chose based on their personality test. There's also the issue that many assumptions about personalities don't necessarily hold true. For example Tal was well known for his tactical wizardry but isn't as well known for his endgames that he often played really well. In the same way Kramnik who is well known for his positional play has played some tactical games really well. So as you get stronger the concept of universal players does get more prominent. Now as far as my suggestion for how to get a chess personality I would suggest the following approaches. 1. Sharpness of positions reached is probably a good indication of how tactical a position is. This can be done with Leela's WDL score with people taking more risks having a higher sharpness score. 2. For technical skills I think endgame accuracy could be a good metric. For example if you use 7 pieces for cut off to endgames and compare it to tablebase I think that might be useful as a metric potentially then modified for rating. 3. For things like "Wildcard" and "Mad Scientist" you could use opening choice and see how often they play lines that aren't frequently used. If by move 5 or 10 they're in completely offbeat lines that would give a higher score there. Players like Basman would score really high there. 4. I'm unsure if it's easy to see but if possible seeing where players win the games. If it's in opening then they're theoretical and do opening prep. If it's in middlegame then it's middlegame/attacking skills and if it's in endgame then they're technical players. 5. Initiative/Aggression I think can be measured by non-material engine evaluations. For example if engine says I'm -0.5 but I'm a pawn down then I'm +0.5 in initiative and my opponent is -0.5 on initiative but +1.0 in material so would get a materialistic trait instead. Players who regularly get a +score excluding material would tend to be attacking players while those with a negative score I think tend to be defensive players. I would imagine most common positions would be ones where one side is +0.3 and has a slight edge but it's drawn with best play. This would also capture sacrifices to some extent. Hopefully that's useful for providing metrics

Absolute cinema

Absolute cinema

@KMcGeoch said in #4:

"Are you a player who is willing to forego castling in favor of a strong attach?" - attach should be attack if you want to fix typo

Thanks - fixed this.

Actually as far as forced surveys go I think the problem that can sometimes occur is that sometimes people don't answer honestly but answer to get the result they want. For example if you identified as Belle as a Disney princess and were asked whether you'd consider dating someone who is abusive and on the hairy side you might answer yes to get Belle. A funny story happened when I knew someone taking a blind taste test for cokes and was a big Coca-cola fan so they gave one that they thought was Coca-cola 10/10 and rest got a 0/10 and then ruthless mocked them when they gave 10/10 to a different brand.

Yes - that's a known problem with self-report measures. Depending on the scale there are some ways to try and mitigate the problem, but that's an argument in favor of behavioral measures like what you can get out of a long-term record of play.

Actually for openings I've had reverse happen where people try to match my "Chess Personality" to get openings and suggest what to play based on that. So potentially someone could get do a quiz to get a personality, choose an opening based on it and then later you base their personality on the opening they chose based on their personality test.

I think there's a question there about whether people would stick to an opening repertoire if it didn't suit them, but I see your point about the possible self-fulfilling prophecy here.

There's also the issue that many assumptions about personalities don't necessarily hold true. For example Tal was well known for his tactical wizardry but isn't as well known for his endgames that he often played really well. In the same way Kramnik who is well known for his positional play has played some tactical games really well. So as you get stronger the concept of universal players does get more prominent.

Yes, I think the question of whether modern GMs are kind of converging on an optimal approach to the game rather than displaying their own "style" per se. It might be neat to look at 960 or short time controls to see if this looks different.

I think sharpness is an interesting aspect of chess positions for a lot of reasons, so I appreciate that suggestion. The other big question behind some of your ideas is whether (and why) we should accept traits like aggression, etc. at all! Thanks again for the great comments!

@KMcGeoch said in #4: > "Are you a player who is willing to forego castling in favor of a strong attach?" - attach should be attack if you want to fix typo > Thanks - fixed this. > Actually as far as forced surveys go I think the problem that can sometimes occur is that sometimes people don't answer honestly but answer to get the result they want. For example if you identified as Belle as a Disney princess and were asked whether you'd consider dating someone who is abusive and on the hairy side you might answer yes to get Belle. A funny story happened when I knew someone taking a blind taste test for cokes and was a big Coca-cola fan so they gave one that they thought was Coca-cola 10/10 and rest got a 0/10 and then ruthless mocked them when they gave 10/10 to a different brand. > Yes - that's a known problem with self-report measures. Depending on the scale there are some ways to try and mitigate the problem, but that's an argument in favor of behavioral measures like what you can get out of a long-term record of play. > Actually for openings I've had reverse happen where people try to match my "Chess Personality" to get openings and suggest what to play based on that. So potentially someone could get do a quiz to get a personality, choose an opening based on it and then later you base their personality on the opening they chose based on their personality test. I think there's a question there about whether people would stick to an opening repertoire if it didn't suit them, but I see your point about the possible self-fulfilling prophecy here. > > There's also the issue that many assumptions about personalities don't necessarily hold true. For example Tal was well known for his tactical wizardry but isn't as well known for his endgames that he often played really well. In the same way Kramnik who is well known for his positional play has played some tactical games really well. So as you get stronger the concept of universal players does get more prominent. Yes, I think the question of whether modern GMs are kind of converging on an optimal approach to the game rather than displaying their own "style" per se. It might be neat to look at 960 or short time controls to see if this looks different. I think sharpness is an interesting aspect of chess positions for a lot of reasons, so I appreciate that suggestion. The other big question behind some of your ideas is whether (and why) we should accept traits like aggression, etc. at all! Thanks again for the great comments!

@NDpatzer said in #6:

Yes, I think the question of whether modern GMs are kind of converging on an optimal approach to the game rather than displaying their own "style" per se. It might be neat to look at 960 or short time controls to see if this looks different.

I think sharpness is an interesting aspect of chess positions for a lot of reasons, so I appreciate that suggestion. The other big question behind some of your ideas is whether (and why) we should accept traits like aggression, etc. at all! Thanks again for the great comments!

Well as far as modern GMs go it even applies to weaker players like myself. I used to not be very tactical and while solid would rely on endgame skills for a lot of my wins and converted small advantages. Then I came to the conclusion that if I wanted to improve then I'd need to work on my weaknesses since those errors are ones that cost me games and so spent a lot of time on that and I now can handle tactical positions much better and don't miss tactics I did before. My reasoning was that before I was a one trick pony and now I have extra tools I can use to win games since I can handle a much broader range of positions. Whether my chess personality changed when I gained 500-600 elo or not is a hard question to answer.

Well I assumed that aggression is important since for example a "Caveman" personality would tend to launch their h pawn up or play f4/Rf3/Rh3 etc and go for a quick mate. On the other hand something like an "Anaconda" personality would be more likely to have positional binds and gradually squeeze the opponent. Players like Petrosian if anything were known for his defensive skills.

You may find players who are fond of gambits that will always find a way to generate a strong attack at a cost of a pawn or two where if it works they get a nice win and if it fails they'll almost always lose the endgame.

For personalities I've assumed you get the following diametrically opposed traits

  1. Positional vs Tactical
  2. Intuition vs Calculation
  3. Sharpness vs Solidity
  4. Aggressive vs Passive
  5. Material vs Initiative/Activity
  6. Theoretical/Technical vs Practical

It's not perfect but I think it broadly classes how most players are. I'm sure you know some players games are always entertaining while others are often quite dull and likely to lead to a draw.

@NDpatzer said in #6: > Yes, I think the question of whether modern GMs are kind of converging on an optimal approach to the game rather than displaying their own "style" per se. It might be neat to look at 960 or short time controls to see if this looks different. > > I think sharpness is an interesting aspect of chess positions for a lot of reasons, so I appreciate that suggestion. The other big question behind some of your ideas is whether (and why) we should accept traits like aggression, etc. at all! Thanks again for the great comments! Well as far as modern GMs go it even applies to weaker players like myself. I used to not be very tactical and while solid would rely on endgame skills for a lot of my wins and converted small advantages. Then I came to the conclusion that if I wanted to improve then I'd need to work on my weaknesses since those errors are ones that cost me games and so spent a lot of time on that and I now can handle tactical positions much better and don't miss tactics I did before. My reasoning was that before I was a one trick pony and now I have extra tools I can use to win games since I can handle a much broader range of positions. Whether my chess personality changed when I gained 500-600 elo or not is a hard question to answer. Well I assumed that aggression is important since for example a "Caveman" personality would tend to launch their h pawn up or play f4/Rf3/Rh3 etc and go for a quick mate. On the other hand something like an "Anaconda" personality would be more likely to have positional binds and gradually squeeze the opponent. Players like Petrosian if anything were known for his defensive skills. You may find players who are fond of gambits that will always find a way to generate a strong attack at a cost of a pawn or two where if it works they get a nice win and if it fails they'll almost always lose the endgame. For personalities I've assumed you get the following diametrically opposed traits 1. Positional vs Tactical 2. Intuition vs Calculation 3. Sharpness vs Solidity 4. Aggressive vs Passive 5. Material vs Initiative/Activity 6. Theoretical/Technical vs Practical It's not perfect but I think it broadly classes how most players are. I'm sure you know some players games are always entertaining while others are often quite dull and likely to lead to a draw.

@NDpatzer said in #1:

Comments on lichess.org/@/ndpatzer/blog/science-of-chess-what-does-it-mean-to-have-a-chess-personality/we5p4kq7

"Are you a player who is willing to forego castling in favor of a strong attach?" - attach should be attack if you want to fix typo

Actually as far as forced surveys go I think the problem that can sometimes occur is that sometimes people don't answer honestly but answer to get the result they want. For example if you identified as Belle as a Disney princess and were asked whether you'd consider dating someone who is abusive and on the hairy side you might answer yes to get Belle. A funny story happened when I knew someone taking a blind taste test for cokes and was a big Coca-cola fan so they gave one that they thought was Coca-cola 10/10 and rest got a 0/10 and then ruthless mocked them when they gave 10/10 to a different brand.

Actually for openings I've had reverse happen where people try to match my "Chess Personality" to get openings and suggest what to play based on that. So potentially someone could get do a quiz to get a personality, choose an opening based on it and then later you base their personality on the opening they chose based on their personality test.

There's also the issue that many assumptions about personalities don't necessarily hold true. For example Tal was well known for his tactical wizardry but isn't as well known for his endgames that he often played really well. In the same way Kramnik who is well known for his positional play has played some tactical games really well. So as you get stronger the concept of universal players does get more prominent.

Now as far as my suggestion for how to get a chess personality I would suggest the following approaches.

  1. Sharpness of positions reached is probably a good indication of how tactical a position is. This can be done with Leela's WDL score with people taking more risks having a higher sharpness score.
  2. For technical skills I think endgame accuracy could be a good metric. For example if you use 7 pieces for cut off to endgames and compare it to tablebase I think that might be useful as a metric potentially then modified for rating.
  3. For things like "Wildcard" and "Mad Scientist" you could use opening choice and see how often they play lines that aren't frequently used. If by move 5 or 10 they're in completely offbeat lines that would give a higher score there. Players like Basman would score really high there.
  4. I'm unsure if it's easy to see but if possible seeing where players win the games. If it's in opening then they're theoretical and do opening prep. If it's in middlegame then it's middlegame/attacking skills and if it's in endgame then they're technical players.
  5. Initiative/Aggression I think can be measured by non-material engine evaluations. For example if engine says I'm -0.5 but I'm a pawn down then I'm +0.5 in initiative and my opponent is -0.5 on initiative but +1.0 in material so would get a materialistic trait instead. Players who regularly get a +score excluding material would tend to be attacking players while those with a negative score I think tend to be defensive players. I would imagine most common positions would be ones where one side is +0.3 and has a slight edge but it's drawn with best play. This would also capture sacrifices to some extent.

Hopefully that's useful for providing metrics

@NDpatzer said in #1: > Comments on lichess.org/@/ndpatzer/blog/science-of-chess-what-does-it-mean-to-have-a-chess-personality/we5p4kq7 "Are you a player who is willing to forego castling in favor of a strong attach?" - attach should be attack if you want to fix typo Actually as far as forced surveys go I think the problem that can sometimes occur is that sometimes people don't answer honestly but answer to get the result they want. For example if you identified as Belle as a Disney princess and were asked whether you'd consider dating someone who is abusive and on the hairy side you might answer yes to get Belle. A funny story happened when I knew someone taking a blind taste test for cokes and was a big Coca-cola fan so they gave one that they thought was Coca-cola 10/10 and rest got a 0/10 and then ruthless mocked them when they gave 10/10 to a different brand. Actually for openings I've had reverse happen where people try to match my "Chess Personality" to get openings and suggest what to play based on that. So potentially someone could get do a quiz to get a personality, choose an opening based on it and then later you base their personality on the opening they chose based on their personality test. There's also the issue that many assumptions about personalities don't necessarily hold true. For example Tal was well known for his tactical wizardry but isn't as well known for his endgames that he often played really well. In the same way Kramnik who is well known for his positional play has played some tactical games really well. So as you get stronger the concept of universal players does get more prominent. Now as far as my suggestion for how to get a chess personality I would suggest the following approaches. 1. Sharpness of positions reached is probably a good indication of how tactical a position is. This can be done with Leela's WDL score with people taking more risks having a higher sharpness score. 2. For technical skills I think endgame accuracy could be a good metric. For example if you use 7 pieces for cut off to endgames and compare it to tablebase I think that might be useful as a metric potentially then modified for rating. 3. For things like "Wildcard" and "Mad Scientist" you could use opening choice and see how often they play lines that aren't frequently used. If by move 5 or 10 they're in completely offbeat lines that would give a higher score there. Players like Basman would score really high there. 4. I'm unsure if it's easy to see but if possible seeing where players win the games. If it's in opening then they're theoretical and do opening prep. If it's in middlegame then it's middlegame/attacking skills and if it's in endgame then they're technical players. 5. Initiative/Aggression I think can be measured by non-material engine evaluations. For example if engine says I'm -0.5 but I'm a pawn down then I'm +0.5 in initiative and my opponent is -0.5 on initiative but +1.0 in material so would get a materialistic trait instead. Players who regularly get a +score excluding material would tend to be attacking players while those with a negative score I think tend to be defensive players. I would imagine most common positions would be ones where one side is +0.3 and has a slight edge but it's drawn with best play. This would also capture sacrifices to some extent. Hopefully that's useful for providing metrics

@NDpatzer

The 500 elo player who chose to retreat their knight instead of attacking: "Yeah I'm a Karpov-Petrosian hybrid".

Great post - it's hard to think of how to measure chess personality.

The only metrics I can think of are attacking vs positional. Also willingness to simplify to endgame. Also whether they prefer bishop or knight. Also willingness to enter complications. Not sure how to put these disparate entities together.

I dunno how to deal with positional players. It is possible to be bad at attacking, and it is possible to be bad at positional chess. But the thing is, the bad attacker can still be called an 'attacker', however the bad positional player cannot be called a 'positional player'.

Maybe at the higher levels its just one personality: 'great at chess'. As Najdorf once said: "Perfection has no style".

But at the higher level there are attacking players who go for complications (Rapport) and there are positional players who go for the endgame (Sarin).

@NDpatzer The 500 elo player who chose to retreat their knight instead of attacking: "Yeah I'm a Karpov-Petrosian hybrid". Great post - it's hard to think of how to measure chess personality. The only metrics I can think of are attacking vs positional. Also willingness to simplify to endgame. Also whether they prefer bishop or knight. Also willingness to enter complications. Not sure how to put these disparate entities together. I dunno how to deal with positional players. It is possible to be bad at attacking, and it is possible to be bad at positional chess. But the thing is, the bad attacker can still be called an 'attacker', however the bad positional player cannot be called a 'positional player'. Maybe at the higher levels its just one personality: 'great at chess'. As Najdorf once said: "Perfection has no style". But at the higher level there are attacking players who go for complications (Rapport) and there are positional players who go for the endgame (Sarin).

Interesting discussion! Thanks again @NDpatzer for stimulating it with your great post! Reading comments and thinking about the better statistical approaches (as well as their theoretical implications), I came to the (provisional) conclusion that maybe factor analysis might not be the best approach to identify (via exploratory FA) and measure (via confirmatory FA) chess personalities/styles. My reasoning is that these methods generally don't handle well cross-loadings, i.e., situations where items (e.g., one tract/indicator pertinent to chess personalities, like one of the many mentioned in the post and comments) load on several factors (see, e.g., Costello & Osborne, 2005). In this case, this means that, for this approach to be viable, one indicator (e.g., preferring sharp positions, or mainly relying on intuition or calculation in given positions) should be indicative of only one factor (also called latent variable), as shown by the strong covariance with other items loading on the same factor. However, I don't think that's necessarily the case, at least for the naive definition of chess personality that is generally shared by the chess community. It might be more useful to think of different combinations of the same indicators/traits defining different chess personalities, in line with the approach of @PETrSmirnov and @KMcGeoch. With this in mind, I think maybe something more on the lines of cluster analysis might be useful: in this case we would not assume one indicator is only associated to a style/personality profile but we could use a bottom-up approach to identifying common combinations of traits (e.g., one common style could be associated with having a medium level of aggressiveness, a high tendency to relying on intuition, a weak preference for sharp positions etc.). I got curious and took a peek and a couple of studies have already tried something like this (McIlroy-Young et al., 2021; Jayasekara, 2021). In particular, the dissertation by Javasekara (see the link below) looks like quite a big dive into this problem, I would have to take some time reading it to say how close it is to what we where hypothesizing.

Costello, A. B., & Osborne, J. (2005). Best practices in exploratory factor analysis: Four recommendations for getting the most from your analysis. Practical assessment, research, and evaluation, 10(1).

Jayasekara, M. G. P. B. (2021). Classification of chess games and players by styles using game data (Doctoral dissertation). URL: https://dl.ucsc.cmb.ac.lk/jspui/bitstream/123456789/4219/1/2014MCS033.pdf

McIlroy-Young, R., Wang, Y., Sen, S., Kleinberg, J., & Anderson, A. (2021). Detecting individual decision-making style: Exploring behavioral stylometry in chess. Advances in Neural Information Processing Systems, 34, 24482-24497.

Interesting discussion! Thanks again @NDpatzer for stimulating it with your great post! Reading comments and thinking about the better statistical approaches (as well as their theoretical implications), I came to the (provisional) conclusion that maybe factor analysis might not be the best approach to identify (via exploratory FA) and measure (via confirmatory FA) chess personalities/styles. My reasoning is that these methods generally don't handle well cross-loadings, i.e., situations where items (e.g., one tract/indicator pertinent to chess personalities, like one of the many mentioned in the post and comments) load on several factors (see, e.g., Costello & Osborne, 2005). In this case, this means that, for this approach to be viable, one indicator (e.g., preferring sharp positions, or mainly relying on intuition or calculation in given positions) should be indicative of only one factor (also called latent variable), as shown by the strong covariance with other items loading on the same factor. However, I don't think that's necessarily the case, at least for the naive definition of chess personality that is generally shared by the chess community. It might be more useful to think of different combinations of the same indicators/traits defining different chess personalities, in line with the approach of @PETrSmirnov and @KMcGeoch. With this in mind, I think maybe something more on the lines of cluster analysis might be useful: in this case we would not assume one indicator is only associated to a style/personality profile but we could use a bottom-up approach to identifying common combinations of traits (e.g., one common style could be associated with having a medium level of aggressiveness, a high tendency to relying on intuition, a weak preference for sharp positions etc.). I got curious and took a peek and a couple of studies have already tried something like this (McIlroy-Young et al., 2021; Jayasekara, 2021). In particular, the dissertation by Javasekara (see the link below) looks like quite a big dive into this problem, I would have to take some time reading it to say how close it is to what we where hypothesizing. Costello, A. B., & Osborne, J. (2005). Best practices in exploratory factor analysis: Four recommendations for getting the most from your analysis. Practical assessment, research, and evaluation, 10(1). Jayasekara, M. G. P. B. (2021). Classification of chess games and players by styles using game data (Doctoral dissertation). URL: https://dl.ucsc.cmb.ac.lk/jspui/bitstream/123456789/4219/1/2014MCS033.pdf McIlroy-Young, R., Wang, Y., Sen, S., Kleinberg, J., & Anderson, A. (2021). Detecting individual decision-making style: Exploring behavioral stylometry in chess. Advances in Neural Information Processing Systems, 34, 24482-24497.