Open product›
Live context, opportunity research and your decision workflow in one place.
How AssetScreener knows — and where it can be wrong.
Plain-language definitions first. Evidence, calibration and mathematics afterwards. Every score must be traceable and challengeable.
Does change + pricing latency + propagation add information?
The engine is only differentiated if those signals improve on simpler price, sector, momentum and fundamental baselines in point-in-time, out-of-sample and live tests.
Change → Transmission → Fundamental confirmation → Expected reaction → Actual reaction → Market Gap → Asymmetry → Red Team → Frozen thesis → Outcome → Learning
One definition everywhere.
Definitions remain stable across product UI, research, structured content and machine-readable summaries.
Market Gap
The estimated difference between how strongly verified change should affect an asset and how much market pricing appears to have absorbed so far.
Permanent definition linkPricing Latency
The measured delay between a verified change becoming economically relevant and that change being reflected in the affected market.
Permanent definition linkInformation Absorption
A structured estimate of which parts of a thesis are already visible in price, estimates, positioning and consensus expectations.
Permanent definition linkPropagation Path
The evidence-backed route through which one change can affect themes, industries, companies and traded assets.
Permanent definition linkConfidence Anatomy
A component view of evidence quality, causal strength, market confirmation, calibration and unresolved objections. It is not automatically probability of profit.
Permanent definition linkRed Team
The deliberate attempt to disprove a thesis, identify missing evidence and define what would invalidate it.
Permanent definition linkProof Ledger
The permanent point-in-time record of what the system said, which evidence it used and what happened afterwards.
Permanent definition linkFacts, inference and interpretation stay separate.
Source-backed observation
A timestamped claim tied to a primary source and named entity.
Computed estimate
A reproducible output with model version, inputs and uncertainty state.
Research judgment
A clearly labelled human or AssetScreener Research interpretation.
No predictive edge is marketed before the evidence exists.
- Point-in-time integrityOnly information available at the decision timestamp.
- Out-of-sample testingHoldout periods and regimes not used for tuning.
- Negative controlsCompare against simpler baselines and shuffled relationships.
- Live frozen signalsPredictions recorded before the observed outcome.
- Permanent failuresWrong theses remain visible in the Proof Ledger.