Event ≠ structure ≠ cause

APEX separates a sourced event, an observed price pattern, and evidence of a reinforcing mechanism. These research classifications are experimental rules, not backtested probabilities or trade instructions.

Event layer

The current adapter supplies sourced FOMC calendar entries. An entry is scheduled before its release, active for two hours afterward, and past after that. An event window does not confirm an actual-versus-expected surprise. The other eleven event families are catalogued but have no connected event feed.

Structure layer

Eight price-pattern rules use completed five-minute bars: direct and delayed repricing, reversal reclaim, second leg, leader contagion, rotation, false break, and compression release. Structures can be candidates without a scheduled event. Data older than twelve minutes or dated in the future cannot create current candidates.

Displacement is evaluated against the configured instrument playbook using completed bars. The first observed crossing is only the first within the available sample, not proof of the true origin of a move. Same-bar observations are tied. Leader-contagion research looks for an ordered progression across related instruments.

Reversal research requires an initial impulse followed by confirmed movement in the opposite direction. It does not prove anyone was trapped. Rotation uses a limited relative-price proxy; its magnitude is neither a probability nor measured capital flow.

Feedback layer

The six components are position stress, leadership spread, breadth expansion, volatility response, liquidity stress, and second-leg strength. Every component must be available before the coverage checklist receives a score. Price and volatility proxies cannot prove forced flow. Full-market breadth, positioning, and order-book feeds are currently missing, so the overall score remains unavailable.

Ten additional mechanism families require specialized sourced observations: cross-asset transmission, short squeeze, long liquidation, gamma acceleration, gamma pin, liquidity vacuum, hedge release, dash for cash, exhaustion reversal, and volatility expansion with forced-flow evidence. They remain unavailable until those inputs exist. Historical examples are context, not training labels or validation results.

Combined Brain scores

The Brain evaluates displacement, propagation, positioning stress, volatility feedback, liquidity stress and continuation alongside cross-market synchronization. Missing required measurements leave the aggregate unscored. These are uncalibrated research rules; event and structure confidence remain unavailable. The public interface explains measured reasons and missing inputs; implementation parameters are maintained on the server.

An unscheduled event can reach CASCADE_ACTIVE only with a complete fresh equity universe, direction, a qualifying structure, displacement, propagation, continuation, breadth, volatility confirmation, synchronized assets, a high aggregate score and separately sourced forced-flow evidence. A candidate pattern alone cannot do this. The older mechanism checklist below the structure inspector is a coverage checklist, separate from the combined Brain score.

Cascade confirmation

The existing strict APEX_SYNC and instrument cascade gates remain separate. Candidate structures never silently promote a trade state or send a text. Current breadth requirements still block cascade confirmation until a real feed supports it.

Why we avoid causal shortcuts

The SEC's GameStop review distinguished periods of short covering from the buying that sustained the broader rally. A rising stock with high short interest is not enough to prove a covering-driven cascade. SEC staff report.

Cboe's analysis distinguishes gross options activity from net dealer exposure and warns against assuming a large hedging effect from volume alone. Cboe market-impact analysis.