Ayan Jain
ReportIndependent quantitative researcher. I build point-in-time market data and audit backtests; backtest-bias is the open library of measured bias constants behind that work.
backtest-bias
Survivorship and data-integrity checks for financial price panels, built on measured constants rather than folklore. Measured and published: 24% of India's 2015 top-500 is invisible to free data; 25.7% of the 2010 S&P 500 (20.1% for 2015); a survivor filter moves a US backtest +0.4 to +1.0 pp/yr, and four recycled tickers moved it 1.7 pp/yr, more than the bias itself. One-line checks for survivorship holes, ticker recycling, and universe construction. Audience: retail quants who cannot afford institutional survivorship-free data, which in India is nearly everyone who backtests. Everything public and git-timestamped: PyPI (backtest-bias), a public ledger of 143 failed strategies, SSRN paper 7099378, Kaggle dataset + notebook, weekly write-ups.
Maintainer support, 12 months
12 months of maintainer support. Milestones: (1) US calibration constants v1; first release SHIPPED 2026-08-13 (Substack Issue #10 + Kaggle dataset). (2) Ticker-identity toolkit: reanimation + first-bar detectors wired into the library. (3) Open US corporate-death registry, every entry with a source URL; 139-name seed staged. (4) Multi-market death-rate comparison: India + US measured, China leg in design. Quarterly community write-ups throughout.
Full-time research year
Everything in the maintainer plan, done full time for a year, sized for a fund that writes larger cheques: the US corporate-death registry completed with a sourced exit for every invisible S&P name (seed of 139 public), the identity toolkit shipped (check_identity and check_universe went live in v0.2.0), price-level look-ahead checks (v0.3), the India-US-China death-rate comparison finished, and the public datasets (Kaggle + the live auction ledger) hosted and maintained. Quarterly write-ups.