Nice, I thought to do something similar, leveraging on Ai (GPT 4o-mini became very cheap now), but struggling on interface because I use python and web app development under it is far from straightforward
Mass Analysis Website

I don't want to use AI models in my website, favoring more traditional yet equally complex statistics models.

Looks cool!
I'm curious, what is the synthetic line? Is that some kind of simulation?
I think having a prediction of rating is very motivating, and having feedback on that could help me adjust how I am approaching improvement. But I think focusing on making it as actionable as possible would get people signing up.
Having something that could predict The One Thing I should focus on to improve would be great (e.g. pawn structure, a certain kind of tactic, a particular opening, or just stop blundering your queen). It could be equally cool to put that as a different color on the graph (assuming some rate and limit of improvement [perhaps eventually measured]).
I also really like Lichess's radar (or "spider-web") chart that shows performance in different puzzle axis. Having it displayed as an area (and a concentric area for similar players) is a really quick way to convey that information. It'd be nice to see that for my real data on here/anywhere (rather than just my puzzles on one platform).
Hello everybody, I'm working on developing a website that aims to provide analysis of trends in all your games on chess.com. The goal is to offer a tool that tracks trends, patterns, and insights, helping players understand their performance and improve their gameplay. Think of it as a homebrew AimChess or Chess.com’s Insights from someone who played chess very seriously for over a year.
Currently, I’ve developed a predictive rating graph that forecasts your rating over time using complex regression techniques. This graph is a key feature, and I plan to expand the website to include other analytics that will help you spot weaknesses in your game. While the site is still in its early stages of development, and I am a very busy person outside of this project (still in school, working part-time, and programming for a world-class robotics team), I hope to keep up regular improvements to this website as a fun personal project.
I would love to hear your feedback and any ideas for features you'd find useful as I continue developing this tool!
*The graph shows the regression model training over the players history (X: game #, Y: rating). Everything past the dashed line represents the program’s predictions
.