Research Areas

Detection and classification of event variables driving short-horizon repricing in public equities, with explicit separation of headline magnitude from estimated business impact.

• 01event detection across regulatory filings, primary disclosures, and curated secondary sources
• 02catalyst classification by structural type and historical materiality distribution
• 03materiality assessment independent of headline sentiment
• 04source recency and timestamp sensitivity
• 05factual consistency reconciliation across publishers

Formalisation of the divergence between aggregate crowd interpretation and reconciled material content of an event, treated as a friction parameter rather than a directional opinion.

• 01estimation of crowd posterior over event relevance
• 02reconciled posterior under multi-source weighting
• 03separation of sentiment from materiality
• 04identification of stale narratives under low refresh rate
• 05information asymmetry without claims of inside information

BUY, SELL, or HOLD as the output of a constrained classification procedure, jointly with confidence, horizon, and an invalidation surface.

• 01discrete classification output with explicit confidence
• 02horizon selected from a ladder calibrated to catalyst class
• 03risk asymmetry quantification across BUY / SELL / HOLD
• 04invalidation logic as a precondition for publication
• 05downgrading to commentary when invalidation cannot be constructed

Short-horizon reaction dynamics, liquidity replenishment, and flow attribution as inputs to confidence — not as standalone directional inputs.

• 01reaction timing relative to event window
• 02liquidity regime classification
• 03volume confirmation and flow attribution
• 04transient versus durable impact decomposition
• 05volatility response and surface deformation

Construction of a consistency graph over source claims with edge weights determined by publisher reliability, timestamp proximity, and structural independence of derivation.

• 01evidence weighting under multi-source disagreement
• 02publisher quality and historical reliability priors
• 03citation graph reconstruction to penalise derivative aggregation
• 04timestamp sensitivity in fast-moving event windows
• 05explicit reporting of source confidence separately from signal confidence

Constraints, audit surfaces, and limitations on machine-assisted interpretation. Includes drift monitoring, backtest discipline, and explicit non-guarantee statements on forward performance.

• 01explicit enumeration of model limitations
• 02uncertainty reporting alongside every classification
• 03drift monitoring on synthesis components
• 04backtest discipline with no claims of forward guarantee
• 05documented invalidation of methodology when reconciliation fails