I think the basic idea is good, but you should add more features. I do not think anybody will be willing to pay money just to see their avg accuracy compared to all rating ranges. (A python script to do so would be one chatgpt prompt away.)
Also i would check the engine evaluation of your "millions of games from lichess" because i got on middlegame a accuracy of 79.45% whereas the 0-100 category gets on avg 84.5%. (I have to admit that the sample of my last 3 rapid games had very poor middlegames but i would have destroyed everybody below 1000 elo)
But in general i think the idea is good.
Hi everyone,
I developed a personalize chess analytics website (www.knightlyKnowledge.com) that uses the Chess.com and Lichess API to export the last 40 user games, analyze each move with stockfish, and present move accuracies bucketed into openings, middlegame, endgame and overall accuracy. I also analyzed 1 Million chess games from a Lichess database of players rated from 100 - 2700 and calculated the average accuracy per rating bucket. This allows me to display a users data alongside histogram plots of the database. I believe this is a unique and novel way to look at chess analysis and show a user where they need to improve based on their last 40 games. I am working on add additional features such as an opening explorer sunburst fan which shows the opening lines you perform well and the ones you lose more often. Also plan to incorporate lessons and PGN studies.
Try it out and let me know if you have any suggestions. If you are interested in more information, DM me and I can provide a coupon code for a free month of Pro subscription.
Thanks!