I watch a lot of anime. I also spend a lot of time in databases. So when I found a dataset of 10 years of MyAnimeList ratings, genres, and studios — I knew what I had to do.
The Dataset
The Kaggle dataset contains:
- ~17,000 anime titles
- Ratings, member counts, and favorites
- Genre tags and studio information
- Airing dates and episode counts
The Questions
- Which studios consistently produce the highest-rated anime?
- Is there a "golden length" for anime series?
- How have genre trends shifted over the decade?
Surprising Findings
Studio quality is remarkably consistent
Madhouse, Bones, and Wit Studio maintain average ratings above 7.5 across their entire catalogs. Studio Deen... does not. The standard deviation tells the real story: some studios are high-variance bets.
12-13 episodes is the sweet spot
Single-cour anime (12-13 episodes) have the highest average rating. Two-cour shows (24-26) are close behind. Anything over 100 episodes drops significantly — length fatigue is real.
Isekai exploded in 2016
Before 2016, isekai (transported to another world) was a niche genre. After 2016, it accounts for 15% of all new anime. The data doesn't lie — we're living in the isekai era.
The Technical Stack
- PostgreSQL for data storage and analysis (CTEs, window functions, pivot queries)
- Power BI for visualization
- Python for data cleaning and import
What I Learned About Data Storytelling
Numbers without narrative are just noise. The anime dataset taught me that the best visualizations answer a question the viewer didn't know they had. "Which studio should I trust?" is more compelling than "Average rating by studio."
Data viz is not about showing data. It's about showing insight.