Data Collection and analyzation, In this section of the project I focused on automating the proccess of scarping the data from the chess.com public API, so I can use it later in training the models. This was an oppurtunity for me to practice thinking algorithmically of most effictive algorithm, since working with huge amounts of data ( 118k+ games and 21m+ raw pieces of data ).

Data Labeling, While the chess.com API provides a large amount of data, it isn’t labeled. I had to label it myself and then use that labeled data to train the models. For example, I added engine evaluations to each position, linked every move to its timestamps, calculated the time deltas to measure thinking time per move, and included additional chess information necessary for training—such as the number of legal moves, whether the position is check, and whether castling is possible. And, of course, I made sure to enjoy my time in the beautiful city I’m visiting—Seoul!

Machine Learning and Data Visualization, Visualizing some data, playing for a bit with it and training the neural networks ( the fun part 🪄)

Initially, I was pressed for time since the challenge I put for myself allowed only a week of development. I may return to this part of the project later, as it’s both the most productive and the one I’m most excited about.