RNG Integrity Studies

Historical personal studies using browser-collected samples and statistical tooling. They explored candidate distributions and dependence tests; finite samples and non-significant tests cannot certify fairness or prove the absence of exploitable patterns.

AI-assisted personal project. Tools and implementation notes are not claims of independent professional specialization or verified live results.

Stack

Engineering highlights

• Sporty Hero crash game: 15-dimension analysis — distribution fitting (5 models), independence (ACF, Ljung-Box), extreme-value (GEV/GPD), and volatility clustering (ARCH/GARCH).
• Earlier reports included fitted-model outputs and estimated house edges. Those figures have not been independently reproduced for this website update.
• Virtual-soccer collection and calibration experiments require checks of sampling, timestamps, execution assumptions, and held-out results before drawing actionable conclusions.
• Reusable pattern: Playwright collector → SQLite (WAL) → analysis pipeline → reproducible report.