Project RYSE - Elite LoL Performance Analytics

Repository: github.com/Guille1799/ryse-publico
Live app: AnalisisClusterRolyLiga

End-to-end R analytics (ETL, clustering, supervised learning, Shiny) developed for an MSc in Behavioural Data Science. The domain is elite League of Legends performance; the methods—heterogeneous groups, predictive modelling, interpretability—transfer to behavioural and social-impact work where averages hide who needs different support.

Overview

Project RYSE analyzes behavioral and performance patterns in high-elo players (Master, Grandmaster, Challenger) using Riot Games data and an end-to-end CRISP-DM workflow.

The project combines:

• data cleaning and feature engineering,
• clustering by role,
• predictive modeling and interpretability,
• and an interactive dashboard for exploration.

Core scope

• Build a clean analytical dataset from match-level records.
• Create performance KPIs for player profiling.
• Identify role-specific archetypes with unsupervised learning.
• Estimate drivers of win probability.
• Provide a practical scouting and comparison interface through Shiny.

Methodology highlights

• Framework: CRISP-DM (full pipeline from data prep to interpretation).
• Data source: Riot Games API (processed and consolidated in data/).
• Filters:
  • elite tiers only: Master / Grandmaster / Challenger
  • invalid roles removed
  • short matches excluded
• Feature engineering:
  • kda_ajustado
  • economy and vision rates per minute
  • objective-related metrics
  • oci (Objective Control Index)
• Modeling and analysis:
  • K-Means clustering (role-wise profiles)
  • Random Forest for key victory factors
  • ALE/PDP-based interpretability workflow
  • consistency metrics (coefficient of variation)

Dashboard sections

The app includes multiple analysis tabs such as:

• General overview KPIs
• Role and cluster profiles
• Correlation analysis
• Key victory factors and variable impact
• Player consistency diagnostics
• Individual player analysis
• Executive report and key findings

Tech stack

• R
• Shiny
• tidyverse
• ranger
• cluster
• pROC
• ggcorrplot
• iml / pdp

Repository structure

.
|-- app.R
|-- data/
|   |-- ryse_database.csv
|   |-- high_elo_puuids_euw.csv
|   `-- random_sample_test.csv
`-- README.md

Run locally

1. Open the project in RStudio.
2. Install required packages (if missing).
3. Run:

shiny::runApp("app.R")

Notes

• This public repository is focused on reproducible analysis and portfolio presentation.
• Some early-stage artifacts and private planning materials are intentionally excluded.

Author

Guillermo Martin de Oliva Carranza
LinkedIn: guillermo-martin-de-oliva-carranza