Mason Flannery
Portfolio

Currently pursuing a Master's in Statistical Data Science at Texas A&M and with a B.S. in Statistics from BYU with experience in statistical modeling, machine learning, and natural language processing.
I am eager to showcase my expertise and contribute to data-driven solutions.

Predicting Dota 2
Professional Games

My most recent project focusing on predicting the outcome of a given professional match of Dota 2, a MOBA video game. This project combines my passion for data science and my interest in gaming.

What is Dota 2?

Click the button above to learn about the game mechanics germane to the project.

Project Summary

The purpose of this project is to develop a predictive model that accurately forecasts the outcome of professional Dota 2 matches. It also offers unexpected insights into the critical aspects of gameplay that players seeking improvement should focus on. By leveraging machine learning techniques and statistical modeling, the project aims to provide valuable insights into the factors that influence match outcomes in Dota 2.

Problem Statement

What is the most crucial aspect of gameplay?
What are the best predictors for team success?

Dota 2 matches are known for their unpredictable nature, which poses challenges for players and fans seeking to understand the key factors contributing to winning or losing. By developing a predictive model, we can gain insights into the determinants of match outcomes and enhance our understanding of the game dynamics.

Techniques Used

In this project, a combination of data preprocessing, exploratory data analysis, feature engineering, and machine learning modeling techniques were employed. The data preprocessing phase involved collecting fresh data and developing a custom Elo algorithm for team ranking. The raw dataset was cleaned and transformed for further analysis. Exploratory data analysis and domain knowledge helped uncover patterns and relationships within the data. Feature engineering was performed to create meaningful predictors for the model. Machine learning algorithms, specifically random forests and logistic regression, were utilized to build predictive models for match outcomes. Finally, the models were tested against the performance of 18 teams from around the world at the highly competitive Bali Major tournament with a $500,000 prize pool.

Key Outcomes and Insights

Through this project, a predictive model with high accuracy in forecasting Dota 2 match outcomes was successfully developed. The model helped discover various features that can teach players what leads to a win in a given match. Analyzing the model's coefficients provided insights into the relative importance of different factors in determining match outcomes. Additionally, key patterns and trends in the data were identified, shedding light on successful gameplay strategies and dynamics.

Overall, this project not only showcases proficiency in data science techniques but also demonstrates the ability to apply these skills in the context of an engaging and complex domain like Dota 2. It reflects a passion for both data analysis and gaming, and I'm excited to share the project and its insights! Feel free to explore the highlights of the project below!

The Data

More on the Data

How the data was collected, what data was collected, and how I explored the data. Includes problems encountered and how they were overcome.

The Models

More on the Model

Considerations in building the models. How I engineered and selected features. How accurate the model is.

The Author

Me skiing at Deer Valley

About me, Mason Flannery! My resume and personal interests.