Octagon Insights — UFC Analytics
About Project
- Type: Team Analytics Project (End-to-End Pipeline)
- Dataset: 1,000+ UFC fight records with 100+ attributes (fighter stats + fight outcomes)
- Deliverables: Cleaned dataset + EDA notebook + visualization suite + baseline ML models for outcome prediction
Objective
Use exploratory analysis and modeling to uncover patterns correlated with UFC fight outcomes, communicate insights through clear visual storytelling, and build baseline predictive models to test which features matter most.
Tools & Technologies
Python, Pandas, Matplotlib, SQL, Jupyter Notebooks, Feature Engineering, Tree-Based Models (Feature Importance), Model Evaluation
Key Work & Impact
Built a reproducible data pipeline that transforms raw UFC fight data into a cleaned analysis-ready dataset by resolving missing values, inconsistent formats, and outliers across a wide column set.
Performed structured EDA to test relationships between fighter attributes (reach, stance, height/weight) and outcome signals (win/loss, finish round, method), translating findings into coach/analyst-readable visuals.
Created 10+ visualizations to highlight performance patterns, distribution shifts, and matchup factors that correlate with winning—designed for storytelling, not just charts.
Developed baseline machine learning approaches to predict fight outcomes and compare models across iterations, using evaluation metrics to identify what improved performance and what did not.
Ran feature-importance analysis (tree-based models) to rank which fighter attributes contributed most to prediction strength, helping separate “interesting” variables from “useful” ones.
Packaged the project for replication (scripts + notebooks + requirements) so anyone can rerun the pipeline end-to-end with the raw dataset.