In this project, I analyze the 18k and 9d subsets of featurecat's Go dataset to discover how final positions differ between low- and high-ranked amateur Go players. Both of the analyses I perform rely on the covariance matrix of the dataset (hence "Go-Variance"), and we make sense of the analyses through pretty visualizations. Check out writeup.pdf for the whole story! (Download the writeup to make the clickable links in the writeup work.)
The data I used for this project can be found here. To use the data with the program, extract the file into a folder called data inside the folder containing src. The folders with the suffix featurecat contain the original game data in .sgf format. The folders with the suffix csv contain the final board states of the games in the corresponding featurecat folder. Finally, the folders with the suffix norm contain the mean-normalized versions of the csv data.
I also include the plots I generated based on the data. In the covariance folder, you will find the visualization of the correlation matrices for each subset, as well as visualizations of the correlations associated with each of the 361 positions of the board. The latter images are simply the reshaped rows of the former. In the PCA folder, you will find visualizations of the principal components for each dataset.