CSK Performance Trajectory Analysis

Submission for the Analytical Assignment. Question: is Chennai Super Kings improving, stagnating, or declining as an institution? Verdict reached by pre-registered criteria, not by narrative.

Verdict

Stagnating — 2 of 5 pre-registered criteria for "declining" were met. Downward win% trend (−3.5pp/yr since 2018) and a sharp playoff-access drop (91.7% → 33.3% in the expansion era) are offset by intact close-match win% and a small peer-relative gap.

Full analysis: deliverables/analysis_note.md

Repo layout

CSK-PerformanceTrajectoryAnalysis/
├── Kaggle-Dataset/                  Raw downloaded CSVs (do not modify)
├── data/
│   ├── bronze/                      Typed Parquet, no cleaning
│   ├── silver/                      Conformed, canonicalized
│   ├── gold/                        Question-shaped marts (CSV + Parquet)
│   └── reference/                   Hand-curated CSK captains, season standings
├── notebooks/
│   ├── 01_bronze_ingest.ipynb       Raw -> bronze Parquet
│   ├── 02_silver_build.ipynb        Team canonicalization, QC gates
│   ├── 03_gold_marts.ipynb          6 analytical marts
│   └── 04_eda_csk.ipynb             Figures + verdict criteria evaluation
├── deliverables/
│   ├── analysis_note.md             Main written deliverable
│   └── figures/                     PNG outputs from notebook 04
├── src/                             Helper utilities
└── activity.md                      Original assignment brief

Reproducing

pip install pandas pyarrow duckdb matplotlib nbconvert
python -m nbconvert --to notebook --execute notebooks\01_bronze_ingest.ipynb --inplace
python -m nbconvert --to notebook --execute notebooks\02_silver_build.ipynb --inplace
python -m nbconvert --to notebook --execute notebooks\03_gold_marts.ipynb --inplace
python -m nbconvert --to notebook --execute notebooks\04_eda_csk.ipynb --inplace

Run in order. Each notebook reads from the previous layer and writes to the next. End-to-end runtime ~30 seconds on a laptop.

Data lineage

• Primary source: Kaggle IPL ball-by-ball + matches, 2008–2025 (1,169 matches / 278,205 deliveries).
• Hand-curated reference: data/reference/ — CSK captains by season and league standings. See the reference README for sources and rationale.

Method

Medallion architecture (bronze → silver → gold) with explicit quality gates at silver. Verdict criteria were declared in 04_eda_csk.ipynb before any trend chart was drawn, to prevent confirmation bias.