C+

Lab · Pattern Methodology

Pattern Lab.

Methodology home for every formalized pattern in the Closelook Pattern Engine. Each pattern starts here as a Lab Read with its academic foundation, lands in the Engine as live signals, and carries a methodology summary that connects the two.

Pattern Engine — live output →

The dual structure

Closelook organizes pattern research as a deliberate split: Lab explains the why; Engine shows you the now. Lab pages document the academic foundation, the failure modes, the worked historical examples — the full theory. Engine pages publish the live signals, updated every Friday after the US close. Each pattern carries a methodology summary on the Lab side and a live signal table on the Engine side, with both linking back to the long-form Lab Read in the Reports section.

The architecture is hierarchical by design. Regime sits at the top: every pattern below operates against an explicit regime classification, and most are suppressed during regime transitions. Trend sits in the middle: the trend context for each name is established before any pattern signal is allowed to fire. Pattern signals like PEAD sit at the third layer: deterministic, mechanical, designed to combine with the upstream context to produce candidate signals — never standalone trade recommendations.

Closelook Pattern Stack — three layers: regime (top), trend (middle), pattern (bottom).
The three-layer Pattern Stack — regime → trend → pattern.

Pattern roster

  • The 60-year market anomaly that survives. 3-day cumulative abnormal return classification with 63-day holding window.

  • Sector Relative Strength

    Coming

    Rotational rankings of the eleven SPDR sectors against the broader index — momentum and mean-reversion signals across the sector cohort.

  • Regime Signature

    Coming

    Quantitative distinction between Grind-Up and Drift-Down behavioral textures — when continuation patterns work and when they don't.

  • Support Confluence at Regression Channels

    Coming

    Deterministic identification of high-probability support clusters from log-regression-channel boundaries.

  • Rolling Engle-Granger + Hurst + Halflife (Ornstein-Uhlenbeck) + Cross-Correlation Function — applied to BTC/Gold (substitution thesis) and BTC/NDX (lead-lag thesis), plus a 2x2 Macro-Quadrant joint reading.

See the patterns live

The Pattern Engine publishes its weekly output every Friday after the US close. PEAD signals are tagged distinctly with their CAR magnitude, regime context, trend status, the universe scanned, and the expected 63-day hold horizon.