Determinants of Success in UFC Competition: An Analysis of Performance, Style, and Physical Attributes

Author: Tin Trung Nguyen

---

Motivation

As a long-time fan of mixed martial arts and the UFC, I have often wondered what truly determines success at the highest level of competition.

With fighters coming from diverse backgrounds and specializing in different styles, this project explores, from a data-driven perspective, which traits and performance factors contribute most to winning.

Is striking more important than grappling?
Does volume matter more than precision?
Do physical advantages still play a significant role?

This project seeks to answer these questions using historical fight and fighter data.

---

Disclaimer

This project is intended for educational and analytical purposes only. It presents a data-driven exploration of historical UFC performance patterns and is not intended to be used for betting, gambling, or financial decision-making.

---

Research Questions

• What in-fight statistics are most associated with winning?
• How do physical attributes (height, reach, weight, etc) impact outcomes?
• Are there specific styles or traits that dominate the UFC?
• What is the relative importance of striking versus grappling?
• How does age affect fighter performance over time?

---

Dataset

• Source: UFC Fights & Fighter Stats Dataset (Kaggle)
• UFC Fights & Fighter Stats Dataset (Kaggle)
• Contains ~20,000 fights and fighter profiles
• Includes physical attributes, performance statistics, and fight outcomes

---

Data Processing

Key preprocessing steps:

• Removed incomplete and unusable fight records
• Cleaned physical attributes (height, weight, reach) using domain knowledge and regression-based imputation
• Split combined in-fight statistics into separate numerical features
• Grouped fight outcome methods into semantic categories
• Created separate datasets for age-based and performance-based analysis

---

Methodology

• Exploratory Data Analysis (EDA)
• Feature engineering
• Statistical analysis
• Logistic regression modeling
• Visualisation using matplotlib, seaborn

---

Key Findings

• Technical skill is the strongest determinant of success. Performance-based metrics such as striking output, striking defence, and takedown defence show the clearest separation between winners and losers.

• Physical attributes have limited impact. Height, weight, and reach display weak associations with fight outcomes, suggesting that physical advantages alone are insufficient at the elite level.

• Striking performance is more predictive than grappling metrics. Composite analysis indicates that striking efficiency and defensive effectiveness have a stronger relationship with winning than grappling statistics.

• Younger fighters hold a modest advantage. Winners are, on average, approximately 0.8 years younger than their opponents, indicating that relative youth provides a small but consistent edge.

• Well-rounded fighters perform best. Cluster analysis reveals distinct fighting styles, with balanced, defensively sound fighters achieving higher win rates than highly specialised profiles.

---

Tools Used

• Python (Pandas, NumPy, Scikit-learn)
• Jupyter Notebook
• Git & GitHub

---

Project Structure

data/ raw/ processed/ notebooks/ data-processing.ipynb README.md

Requirements

This project was developed using the following tools and libraries:

• Python 3.8+
• Jupyter Notebook
• pandas
• numpy
• matplotlib
• scikit-learn
• seaborn

The main analysis is contained in the Jupyter notebook and can be run in any standard Python environment with these dependencies installed.