Steam Games Exploratory Data Analysis & Dashboard

A Python and Flask project to analyze over 114k games in the Steam Games Dataset. Features a clean dark-mode dashboard with interactive charts (using ApexCharts) and a static fallback mode (Matplotlib).

---

Quick Setup

1. Local Development (Flask App)

This runs the full Python server locally, which loads and processes the dataset in-memory.

1. Install requirements:

   pip install -r requirements.txt

2. Add the data: Grab games.csv from Kaggle and place it in a data/ folder in the root directory.
3. Run it:

   python app.py

   Open http://localhost:5000 in your browser.

Note: You can still run the Jupyter notebook (steam_eda.ipynb) locally for the raw EDA code.

---

The Comma-Shift CSV Bug (Important)

While parsing the Kaggle games.csv file, I found a weird formatting issue in the raw dataset:

• The header line is missing a comma between Discount and DLC count (it reads as DiscountDLC count).
• Because of this, the header line has 39 columns, but the data rows have 40 columns (since they actually have a comma there).
• In previous analyses, this caused pandas to treat column 0 (AppID) as the index, shifting all header names by 1 to the right.
• This shift meant the Price column was actually reading the Discount column! That's why previous stats showed an incorrect median price of $51.00 (which was actually a median discount percentage of 51%).
• It also caused the date-filtering helper to throw out 99.9% of the top games (like CS2, Dota 2, and PUBG) because their game names shifted into the date columns and got parsed as dates.

How it's fixed: I manually defined the correct 40-column headers list in app.py and passed it as names=cols when loading the CSV, completely fixing the column alignment for all 122,611 rows.

---

Corrected Key Metrics

With the CSV parsing fixed, the actual Steam dataset stats are:

• Free-to-Play Ratio: 15.67% (17,902 out of 114,198 games) are actually free-to-play. (The old shifted parsing showed 64% free because it counted any game with "0% discount" as free).
• Median Pricing: The actual median price of paid games is $3.49 (demonstrating the massive tail of low-cost indie games and budget titles on Steam). 90% of paid games in this dataset are under $12.00.
• Price vs. Quality: The correlation between a game's price and its positive rating ratio is near-zero (0.015), showing that price does not indicate better quality.
• Top 15 Most Reviewed: The most-reviewed chart is now correctly populated with global blockbusters (Counter-Strike 2, PUBG, Dota 2, etc.) instead of being empty.

---

Visuals & Frontend

• UI Theme: Clean dark mode inspired by Steam, with glassmorphism panels.
• Metrics count-up: Smooth JS number count-up on load.
• Interactive Mode (ApexCharts): Animated charts with custom hover tooltips showing game name, pricing, and exact ratings.
• Static Mode: A toggle in the header switches the layout to render the original Matplotlib/Seaborn static images.
• Performance: Sub-10ms response times because all data aggregation is cached in memory.