Data & Finance

7 projects. Python, pandas, scikit-learn, Streamlit, financial modeling.

Equity DCF dashboard valuing Apple from its 10-K filings, with a free cash flow chart

Equity DCF

Type any US-listed ticker and get a discounted cash flow valuation built from its audited 10-K filings.

  • Pulls the full 10-K history from the SEC EDGAR XBRL API. No API key and no hand-keyed numbers.
  • Keeps only annual facts with a 330 to 400 day filter, so quarterly numbers never get counted as annual.
  • Outputs a WACC × terminal growth sensitivity grid and a football-field chart against the live market price.
  • Python
  • pandas
  • Plotly
  • Streamlit
  • SEC EDGAR
ROC curves with the random forest (AUC 0.776) above logistic regression (AUC 0.708)

Credit Default Prediction

Predicts credit card default on 30,000 accounts, with a live risk scorer.

0.78ROC-AUC, random forest
30,000credit card accounts
  • Logistic regression vs random forest (0.71 vs 0.78 ROC-AUC) on the UCI dataset, with class weights for the 22% default rate.
  • Recent repayment status was the strongest predictor.
  • The Streamlit scorer comes with notes on fairness and responsible use.
  • Python
  • scikit-learn
  • Streamlit
Bar chart of rating by genre, with indie and strategy at the top

Steam Games Analysis

What makes a Steam game succeed. Python analysis and a live dashboard on the 400 most-owned Steam games.

92% vs 81%positive reviews, paid vs free-to-play
400most-owned Steam games analysed
  • Pulled the 1,000 most-owned games from the SteamSpy API, then added genres and tags for the top 400.
  • Action is the most common genre at 72% of titles, and it is not the highest rated.
  • The dashboard filters by genre, price, and review count.
  • Python
  • pandas
  • Plotly
  • Streamlit
Chart of 32GB DDR5 kit prices since 2023 with fast, base and slow forecasts falling back to $111

RAM Price Outlook

When RAM, SSD, GPU and CPU prices get back to where they were before the AI-driven memory shortage.

4.9x32GB DDR5 kit now vs normal ($540 vs $111)
Dec 2030RAM back to normal, base case
  • Every number is computed in the repo from free public data, including government price indices, company filings, archived retail prices, and market data.
  • Each number is marked measured (read from a source) or modeled (output of the forecast).
  • RAM gets back to normal between Aug 2029 and Jul 2031 across the fast and slow cases. 2TB NVMe SSDs are at 2.4x normal.
  • Python
  • pandas
  • Forecasting
¥3.97bncapex
¥322Mstabilised revenue a year
GOas a public-private partnership

Shizuoka Ice Arena Feasibility Study

Private engagement

Feasibility study for a junior-Olympic ice skating and ice hockey arena in Shizuoka Prefecture, Japan.

  • Shizuoka has 3.55M residents and no indoor rink since the Hamamatsu Sports Center closed in May 2024.
  • Recommended GO as a public-private partnership in Hamamatsu, with about +¥50M a year operating surplus under a private operator.
  • 18 documents covering the demand model, capex stack, revenue build, and operator economics, built in 1 day.
  • Financial modeling
  • Market research
  • Feasibility
Idle game market report showing 3,479 games analysed and 8 themes

Idle Game Market Analysis

A data map of the idle and incremental game market on Steam, published as an interactive report.

3,479idle and incremental games analysed
8themes, with nature the underserved niche
  • Seeds the catalogue from SteamSpy's Idler, Clicker and Incremental tags, then adds price, reviews, release dates and update history from the Steam APIs.
  • 9 analysis modules (topics, performance, gaps, quality, monetization, art, updates, trends, naming) feed a static Plotly report.
  • itch.io and Kongregate add context but stay out of the Steam comparison tables, because their data is thinner.
  • Python
  • pandas
  • Plotly
  • Jinja2
12chapters
94 to 99%Steam rating for walk-plus-a-system games, vs 67 to 77% walk-only
$25k to $35kmedian year-one Steam net at $14.99

Hiking Game Research

A 12-chapter market, design and feasibility report on hiking and nature games.

  • Every load-bearing number is graded inline as confirmed, reported, or estimate, and each chapter lists what could not be verified.
  • A $14.99 indie hiking game models to about $25k to $35k median year-one Steam net.
  • Ranks 3 open lanes, and a competitor check found no announced direct rival in the top one (hiking with a seasonal life-sim).
  • Market research
  • Steam data
  • Game design