Financial Econometrics

ARGOS.

Two-asset cointegration as a falsifiable hypothesis: discover the pair, engineer the spread, separate trading policy from selection, and validate under costs, stability, multiplicity, and causal out-of-sample.

Flagship — ARGOS

Two-Asset Cointegration & Statistical Arbitrage Engineering

The object is (A, B): a bivariate relation turned into a falsifiable hypothesis, then into a statistical-arbitrage strategy that survives costs, stability, multiplicity, and causal out-of-sample. Not a high-correlation z-score recipe.

What ARGOS provides

Six technical phases — correlation ≠ cointegration, and OOS ≠ untouched OOS.

Phase 1 — RELATION Two-asset foundations and integration tests I(0) / I(1).

Phase 2 — SPREAD Construction + mean reversion — Spread = A − βB, hedge ratio, and OU diagnostics.

Phase 3 — POLICY Signal, Ornstein-Uhlenbeck, and friction — costs, commissions, and borrow.

Phase 4 — TIME Stability, dynamic tracking, and structural breaks — a broken equilibrium is retired.

Phase 5 — SEARCH Large-scale pair search, multiple testing, and causal out-of-sample validation.

Phase 6 — LAB A reproducible pair-discovery pipeline, failure taxonomy, and capstone engine.