Hello,
As a person that does data analysis for a living, I was recently thinking about how to utilise chess insights. Some of the statistics e.g. average centipawn loss per game phase could influence how to set up training, over a large enough sample you can track your performance. However, in order to gain additional insight, would it be possible to have for all of these statistics 1) you own results, but also 2) a distribution of results from other players in a specified rating range? For example I can see that my average centipawn loss is much lower in opening than middlegame, however it's clear that it's also, because in a natural way you would expect that in opening it's easier to not make blunders. What would be useful is to know the performance with similar filter of my peers, that way we could benchmark the performance against what is expected. I could imagine that such a solution might not be trivial mostly due to database size, but I'm interested in the possibility of that nevertheless.
As a person that does data analysis for a living, I was recently thinking about how to utilise chess insights. Some of the statistics e.g. average centipawn loss per game phase could influence how to set up training, over a large enough sample you can track your performance. However, in order to gain additional insight, would it be possible to have for all of these statistics 1) you own results, but also 2) a distribution of results from other players in a specified rating range? For example I can see that my average centipawn loss is much lower in opening than middlegame, however it's clear that it's also, because in a natural way you would expect that in opening it's easier to not make blunders. What would be useful is to know the performance with similar filter of my peers, that way we could benchmark the performance against what is expected. I could imagine that such a solution might not be trivial mostly due to database size, but I'm interested in the possibility of that nevertheless.
