BasketXplainer

Team Members

1. Dustin Brunner
2. Jonathan Koch
3. Liule Yang
4. Timothé Laborie

Contribution Statement

Every team member contributed approximately equally to the project. Below we list the contributions of each member in descending order of importance:

• Dustin Brunner: Backend, Research, Report, Frontend
• Jonathan Koch: Wireframe, Design, Backend, Tutorial, Report
• Liule Yang: Backend, Research, Report
• Timothé Laborie: Main contributor to frontend (Implementation and Styling)

Deployment Link

Try out our deployed dashboard here

How to Run Locally

Requirements

In this project we are using LightGBM for the ML pipeline. This package can be a bit tricky to install depending on the OS. If the usual installation steps as described below fail, more information for troubleshooting can be found here.

Installation Instructions

To run this project you have to:

• clone the repository;
• open a new terminal instance;
• move to the folder where the project has been downloaded using the command cd;
• open the folder called "b5-winning-in-basketball";
• create a virtual environment from the environment.yml file using the command conda env create --file environment.yml;
• activate the virtual environment run the command conda activate b5-winning-in-basketball; To run the backend
• open the backend folder called "backend-project";
• install the requirements using the command pip install .. If you plan on editing the requirements, you may want to use the command pip install -e .;
• start the backend with the command start-server; To run the frontend
• open a new terminal instance and once again go to the folder called "b5-winning-in-basketball"
• open the frontend folder called "react-frontend";
• start the front end by using the following two commands npm install, npm start; If all the steps have been successfully executed a new browser window will open automatically.

Project Description

The goal of our project is to create an interactive dashboard that allows basketball coaches/analysts to determine the most important factors for predicting the winning odds of a given matchup.

Users

• Basketball Data Analysts (specifically Fran Camba Rodriguez of the Obradoiro CAB team)
• Basketball Coaches

Datasets

Add here all used datasets.
Document here where to find the data and how to download it.

• NBA Kaggle Dataset

Tasks

Define all the tasks you want your dashboard solve.

• Determine feature importance for predicting winning odds
• Modifying boxscore data for interventional predictions (what-if analysis)
• Positioning teams in comparison to other teams according to defensive and offensive performance

Dashboard Structure

Components

• Introductory Tutorial
  • Overlay when first launching the app
  • Help button in the corner
• Team Selector
  • Team logos
  • Scrollable dropdown (search)
• Interactive Box Score Statistics
  • Parallel Coordinates Plot (adjustable ordering)
  • Adjust Box Scores by sliding directly in plot (direct manipulation)
  • Provide some realistic constraints for the box score values
• Winning Odds Prediction
  • If not dynamically updated, indicate the change in winning odds after recalculating
• Explainability Plot
  • SHAP force plot to better comprehend prediction of model
  • Indicate which features contribute to increasing / decreasing the winning odds of the home team
• League Overview
  • Plot the different teams in the league based on offensive and defensive capabilities
  • On-Hover details

---

Folder Structure

Specify here the structure of you code and comment what the most important files contain

├── README.md  # project readme
├── backend-project
│   ├── README.md
│   ├── setup.py   # main app
│   ├── pyproject.toml
│   ├── src
│   │   ├── dummy_server
│   │   │     ├── router
│   │   │     │    ├── routes.py  # api endpoint routes
│   │   │     │    ├── app.py
│   │   │     │    └── __init__.py
│   │   │     └── resources
│   │   │         ├── clustering.py  # league overview endpoint
│   │   │         ├── explainability.py  # feature importance endpoint
│   │   │         ├── games_data.py  # box score endpoints
│   │   │         ├── prediction.py  # prediction endpoint
│   │   │         ├── utils.py  # utility functions
│   │   │         └── __init__.py
│   │   └── __init__.py 
│   ├── data
│   │   ├── precomputed  # precomputed models & box scores
│   │   ├── team_logos  # NBA team logos
│   │   ├── dataset_games.csv  # NBA Kaggle datasets
│   │   ├── dataset_games_details.csv
│   │   ├── dataset_players.csv
│   │   ├── dataset_ranking.csv
│   │   └── generate_teams.csv
│   └── MANIFEST.in
├── dev_notebooks  # development notebooks (including model training)
├── react-frontend
│   ├── README.md
│   ├── Dockerfile
│   ├── package-lock.json
│   ├── package.json
│   ├── src
│   │   ├── App.css
│   │   ├── App.test.tsx
│   │   ├── App.tsx  # main frontend code
│   │   ├── router
│   │   │   ├── resources
│   │   │   │   └── data.ts
│   │   │   └── apiClient.ts
│   │   ├── components
│   │   │   ├── utils.ts
│   │   │   ├── ParallelCoordinates.tsx  # parallel coordinates component
│   │   ├── index.css
│   │   ├── index.tsx
│   │   ├── logo.svg
│   │   ├── react-app-env.d.ts
│   │   ├── reportWebVitals.ts
│   │   ├── setupTests.ts
│   │   └── types
│   │       ├── margin.ts
│   │       └── data.ts
│   ├── tsconfig.json
│   └── public
│        ├── robot.txt
│        ├── manifest.json
│        ├── logo512.png
│        ├── logo192.png
│        ├── index.html
│        └── favicon.ico
├── environment.yml  # package dependencies
└── report.pdf  # project report

Versioning

Create stable versions of your code each week by using gitlab tags.
Take a look at Gitlab Tags for more details.

Then list here the weekly tags.
We will evaluate your code every week, based on the corresponding version.

Tags:

• Week 1: Week 1 Tag
• Week 5: Week 5 Tag