HIP HOOP 🎤🏀

https://nikhil-thomas-a.github.io/hip-hoop/

Does a hip-hop mention change how an NBA player performs?

A full end-to-end data science project analyzing the statistical relationship between hip-hop cultural mentions and NBA player performance. Real game logs. Real lyrics. Real math.

---

Project Structure

hip-hoop/
├── data/
│   ├── raw/
│   │   ├── mentions.csv          # 110 hip-hop NBA mentions with release dates
│   │   └── game_logs/            # per-player game log CSVs (generated, gitignored)
│   └── processed/
│       ├── mentions_with_sentiment.csv
│       ├── windows.csv           # before/after stat windows per mention
│       ├── normalized.csv        # pace-adjusted, era-normalized
│       └── results.json          # ← final output read by index.html
├── notebooks/
│   ├── 01_data_collection.ipynb  # NBA API + Genius API pulls
│   ├── 02_cleaning.ipynb         # sentiment, windows, normalization
│   ├── 03_eda.ipynb              # exploratory analysis + charts
│   ├── 04_statistics.ipynb       # t-tests, correlations, regression
│   └── 05_export.ipynb           # generates results.json
├── src/
│   ├── fetch_gamelogs.py         # nba_api wrapper
│   ├── fetch_lyrics.py           # Genius API fetcher
│   ├── sentiment.py              # VADER sentiment scoring
│   ├── build_windows.py          # before/after window construction
│   ├── normalize.py              # pace + era normalization
│   └── export.py                 # builds results.json
├── index.html                    # interactive executive summary
├── generate_notebooks.py         # regenerates all notebooks from source
├── requirements.txt
├── .env.example                  # copy to .env and add your API key
└── .gitignore

---

Setup

1. Clone and install

git clone https://github.com/yourusername/hip-hoop.git
cd hip-hoop
pip install -r requirements.txt

2. Add your Genius API key

cp .env.example .env
# Edit .env and add your GENIUS_API_KEY
# Get a free key at: https://genius.com/api-clients

⚠️ Never commit your .env file. It is in .gitignore.

3. Run the notebooks in order

jupyter notebook

Open and run each notebook in sequence:

Notebook	What it does	Runtime
01_data_collection	Fetches NBA game logs + Genius metadata	~20 min
02_cleaning	Sentiment scoring, windows, normalization	~2 min
03_eda	Exploratory charts and distributions	~1 min
04_statistics	t-tests, correlations, regression	~1 min
05_export	Generates data/processed/results.json	~30 sec

4. Open the dashboard

# Option A: open directly in browser
open index.html

# Option B: serve locally (avoids fetch() CORS issues)
python3 -m http.server 8080
# then visit http://localhost:8080

---

Methodology

Data Sources

• NBA game logs — nba_api library, pulling per-game stats (PTS, AST, REB) for each player
• Release dates — from mentions.csv, cross-referenced with Genius API
• Sentiment — VADER applied to each lyric snippet

Statistical Windows

For each mention, we construct 5 windows relative to the song release date:

• Baseline: 30 games before
• After 1g: next 1 game
• After 10g: next ~10 games (~2 weeks)
• After 30g: next ~30 games (~1 month)
• After season: remainder of that NBA season

Normalization

• Pace adjustment: scales stats to a 2010s reference pace (NBA pace has varied from ~89 to ~100 possessions/game across eras)
• Era z-scores: standardizes deltas within each era group

Statistical Tests

• Paired t-tests: before vs. after for PTS, AST, REB
• Mann-Whitney U: compliment vs. diss group comparison
• Pearson correlation: artist tier and VADER score vs. performance delta
• OLS regression: composite delta ~ artist tier + mention type + sentiment

Composite Impact Score

PTS delta (40%) + AST delta (30%) + REB delta (30%) — pace-adjusted, 30-game window.

---

Key Research Questions

1. Does a hip-hop mention significantly change performance? (paired t-test)
2. Does mention type (compliment vs. diss) predict direction? (Mann-Whitney U)
3. Does artist tier correlate with magnitude of change? (Pearson r)
4. Does continuous sentiment score outperform binary classification? (regression)

---

Dataset: 110 Mentions

Spanning 1992–2022, covering players from Shaq and MJ in the 1990s to Luka, Ja Morant, and Anthony Edwards in the 2020s. Artists from Notorious B.I.G., Jay-Z, and Nas through Drake, Kendrick Lamar, and Travis Scott.

---

Limitations

This is a correlational study. We cannot claim causation — basketball performance is driven by hundreds of confounding factors (opponent strength, injury status, rest, home/away). We report p-values and effect sizes honestly and acknowledge when results are not statistically significant. The dataset of ~110 mentions is appropriate for exploratory analysis but underpowered for strong causal claims.

Selection bias: we only captured famous, well-documented mentions. Unknown references are underrepresented.

---

Live Demo

Deployed via GitHub Pages: https://yourusername.github.io/hip-hoop

Note: GitHub Pages serves static files. The results.json in data/processed/ must be committed for the live dashboard to work.

---

License

MIT — fork it, remix it, cite it.

---

Built for culture.