Signal · Pattern Engine

Pattern Engine.

Five live pattern surfaces. Each one is the operational view of a single pattern — sorted for fast scan, not deep methodology. Tap a tile for today's signal; tap the methodology chip for how the pattern is computed.

How the engines fit together

The Pattern Library has three surfaces per pattern: the Lab Read is the essay-form thesis (where one exists); the Methodology is the formal definition with formulas; the Engine is the operational scan — what's actionable today.

Sector RS, Regime Signature, and Support Confluence share the same universe (11 SPDR sector ETFs + SPY) and the same nightly EODHD refresh — they read the same JSON snapshot from different angles. Stack them: a name sitting in Grind-Up (regime) with the highest confluence-confidence (support) and at the top of the RS rank (rotation) is the cleanest setup the framework currently produces.

Pattern 03 and PEAD are event-driven instead of price-driven — Pattern 03 fires on NVIDIA disclosures, PEAD on earnings prints. They run on separate data pipelines but share the same dual-structure architecture.

Diary, not advice. Lab patterns refreshed nightly at 22:30 UTC by lab-data-refresh.yml; Support Confluence by the closelook-pattern-scanner worker at 23:05 UTC Mon–Fri.

What the Pattern Engine scans

The engine runs Closelook's pattern library over the full covered universe after each US close — support-confluence at regression channels, sector relative strength, regime signatures, post-earnings drift (PEAD) and the NVIDIA strategic-authority pattern. Each scan page lists the names currently matching, with hit scores and the levels that produced the match. Patterns are detection rules, not forecasts: a hit says the setup EXISTS, the methodology page for each pattern says what has historically followed and how often. The construction, filters and failure modes live in the Pattern Lab; readings here are our own research diary, not investment advice.