Back to search

CSUniversity San Bernardino

STATISTICAL ANALYSIS AND MACHINE LEARNING TO IMPROVE LEAGUE CHAMPIONSHIP SERIES TEAMS

Abstract

dc:description.abstract

ABSTRACT One area for further study in Esports is the use of advanced analytics from a performance standpoint. This culminating experience project sought to find and implement effective performance analytics techniques, using the most popular Esport (League of Legends) as its subject. The research questions asked are (Q1) How do champions, players, and their associated in-game variables impact the results of League of Legends matches? (Q2) How can machine learning algorithms be implemented to utilize descriptive and predictive analytics for League of Legends most effectively? Additionally, while not an element of the analysis and machine learning model, it is important to discern the importance and scope of the data collected. The findings are: (Q1) In game variables can be utilized to create descriptive analytics metrics like Champion Matchup Value (CMV), Composition Pace Factor (CPF), or Overall Pace Rating (OPR), and (Q2). The results from the machine learning model focused on correlation and weighting variables, in conjunction with the metrics formulated from answering Q2 can effectively determine a team’s chance of winning with an associated confidence rating for that prediction. Additionally, the data collected from OraclesElixir presented a broad set of variables and a substantial number of observations that opened the path to more meaningful analysis than in prior studies. The machine learning model, when fed professional matches of League of Legends saw nearly a 70% accuracy rating, with a confidence band to determine the likelihood of outcome in each match. Breaking down the basic statistics into more refined metrics, like CPF or CMV, provides an additional vector from which the game can be analyzed. While some studies aim to use the in-game statistics as they are found, emulating the process of sprots could greatly benefit the world of esports. Creation of advanced analytics allows for a heightened look into how these stats impact games. Additionally, factoring these advanced statistics into a machine learning model which can intake raw in-game statistics, calculate these stats, and utilize them to predict a winner of a match is also beneficial. Many factors come into play with sports, and winners can never be predicted with 100% accuracy, but 70% accuracy is fairly impressive for a model built on original algorithms following brand new advanced statistics. The model also presents a band for how often a certain XFW score results in a winner, thus allowing for more confident predictions with either a high or low XFW score. This model could be further improved to hopefully allow for even more accurate predictions. Areas for further study would range from data collection method to further expanding on this ML model or building a similar model that explores even more variables. There is more game data available from Riot’s API, and even some variables found in the online database, like current-patch, were not included in the final version of this machine learning model but would likely serve only to improve the accuracy of the model.

Degree

thesis:*
Name thesis:degree_name
Master of Science in Information Systems and Technology
Level thesis:degree_level
Project
Discipline thesis:degree_discipline
Information and Decision Sciences

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Gilles, Alexander
Contributors dc:contributor
  • Shayo, Conrad

Subjects

dc:subject × 7

Identifiers

dc:identifier.*
Repository record dc:identifier
https://scholarworks.lib.csusb.edu/etd/1793

Chain of custody

source
Harvested from
CSUniversity San Bernardino
Base URL
scholarworks.lib.csusb.edu/do/oai/
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Gilles, Alexander. STATISTICAL ANALYSIS AND MACHINE LEARNING TO IMPROVE LEAGUE CHAMPIONSHIP SERIES TEAMS. Project thesis, https://scholarworks.lib.csusb.edu/etd/1793