Date of Submission
Campus Only Senior Thesis
Bachelor of Arts
This paper tests the dynamics and interactions of market efficiency in non-traditional markets through the analysis of the virtual Counter Strike Global Offensive Market. The emphasis and implementation of advanced feature generative and reduction processes in Long Short Term Memory Neural Networks are used to not only further test these models’ ability to identify, capture, and predict, but I propose as a potential method to identify and study market interactions and dynamics in nontraditional markets. The framework utilizing Principal Component Analysis along with a Random Forest Regression was taken from best practices outlined in literature for overall model accuracy. Both the PCA-LSTM and RF-LSTM were tested and benchmarked with the traditional model performance metrics of MAE, MSE, and MAPE. These performance metrics were then compared to the performance metrics of a Naive Algorithm and Random Walk. The results found that both LSTM models outperformed the benchmarks with accuracy and differences at statistically significant levels. These results challenge the validity of EMH in digital markets and show the potential for the application of advanced models in enhancing the understanding of market efficiency and dynamics in non-traditional environments.
Ogasawara, Peyton, "Market Dynamics and Interactions: A Study of the Counter-Strike: Global Offensive Virtual Market Through Long Short-Term Memory Neural Networks" (2024). CMC Senior Theses. 3524.
This thesis is restricted to the Claremont Colleges current faculty, students, and staff.