Systematic Equity Research Platform

Alexander Hines · Systematic Equity Research Platform

Signal → point-in-time data → construction → execution, as one loop.

An independent, end-to-end quantitative equity platform: billions of rows of point-in-time market data, a fenced signal-research program, and a paper-execution lane that measures its own fills against the official market print. Its defining discipline is adversarial self-invalidation — every headline result is assumed wrong until it survives leakage, selection, cost, and forward tests.

Paper only · no live capital As of

The honest performance ladder

Generated
1.66Sharpe

Candidate META book (26.1% CAGR) before modeled implementation costs — the ceiling the raw search produced.

Audited → suspended

A look-ahead leak was found in the dominant sleeve — a third of its selections existed only through it. The headline was suspended, not adjusted.

Certified floor
1.40 → 1.05

Pre-registered P1 book: ~1.40 before / 1.05 after modeled costs (1.07 holdout). Clears the registered ≥1.0 bar.

Forward validation

The stronger candidate is under live paper test. Execution cost — not alpha — is the binding uncertainty.

Sharpe figures are annualized, hypothetical/backtested, pre-tax, and shown before vs after conservative modeled implementation costs (trading, spread, borrow). No live-capital track record is claimed. The asset here is the arc — strong generation, rigorous gating, honest survivor — not any single number.

Engineering scale — point-in-time data infrastructure

4.31B

OptionMetrics contract-rows ingested, 0-delta vs source (139 GB)

291.4M

TAQ auction cross-prints, 2003–2026, field-validated

867

point-in-time (bitemporal) data contracts, leak-tested

35,021

catalogued WRDS table-families (115,409 relations mapped)

0

row delta across 4 mid-pull reconnects — reconnect-hardened ingestion

Engineering

Production data infrastructure

Memory-safe, resumable, reconnect-hardened pipelines; a self-generating coverage catalogue; a governed redaction chokepoint behind this very page.

Explore the stack →

Execution

Fills vs the official print

47 real paper fills reconciled field-by-field against official auction crosses; a +116 bp routing defect diagnosed; cost-stressed Sharpe 0.68 → 1.36.

See the cost bridge →

Research

Falsification as method

DSR, purged CV, sealed holdouts, cross-regime IC, pessimistic-cost scoring across ~11,300 recorded looks — and suspension of contaminated 1.2–1.45 headlines.

Read the gates →

Live paper record — forward validation

Paper NAV

Inception

Live Sharpe

Hidden until 20 net-return sessions

Registered anchor

net of 2× stressed cost — a target, not a claim

Drawdown kill line

pre-registered stop
Paper trading only. No real capital is deployed. Holdings are public for this validation; broker account IDs, order IDs, model scores, and secrets are denied at publish time by a governed redaction chokepoint (scripts/ops/dashboard_publish.py).