This document summarizes an unsuccessful experiment using self-organizing maps (SOM) for unsupervised learning on S&P 500 historical index data. The goal was to cluster unusual trading patterns like those during financial crises, but the SOM failed to produce meaningful clusters. Even after adjusting the data set sizes and attributes tracked, the resulting maps showed randomly distributed nodes with no clear separation of clusters. The SOM was only somewhat successful in clustering when tracking a single attribute, but the clusters did not clearly correspond to known unusual periods in the market index.