Scores and Interpretation
Scores help prioritize evidence. They do not convert uncertainty into certainty.
Importance: 1–10
Section titled “Importance: 1–10”Importance estimates how much attention an event deserves relative to other current events.
It can consider:
- materiality of the language or structured event;
- entity relevance;
- source reliability;
- independent corroboration;
- event velocity;
- market or on-chain impact;
- quality of the underlying evidence.
Importance bands can be rendered differently by page. The numeric score is the stable reference; visual labels are presentation aids.
Do not interpret importance as expected return. A major negative event can be highly important, and a highly important event can already be priced in.
Confidence: 0–1 or percentage
Section titled “Confidence: 0–1 or percentage”Confidence estimates the strength of the current evidence.
It can increase with reliable primary sources, data quality, independent corroboration, and consistent evidence. It can decrease with ambiguous entities, stale inputs, unsupported social claims, conflicting reports, or sparse data.
Do not interpret confidence as a calibrated probability that a trade will succeed.
Sentiment
Section titled “Sentiment”Sentiment describes textual tone or direction. It is useful for comparing discussion and detecting change. It is not proof, materiality, or market direction.
Narrative score
Section titled “Narrative score”Narrative score prioritizes developing themes using growth, persistence, breadth, and evidence quality. It is a ranking signal, not a probability that the theme will continue.
Quant scores
Section titled “Quant scores”Quant signals can expose:
- forecast strength;
- expected volatility;
- anomaly score;
- confidence;
- input-data quality.
These scores describe a model output and its inputs. They must be read together. A high anomaly score can indicate unusual behavior without identifying the cause or direction.
Wallet risk: 1–10
Section titled “Wallet risk: 1–10”Wallet risk prioritizes investigation based on the available address profile and labels. It is not a legal judgment or proof of wrongdoing.
Private scores
Section titled “Private scores”Internal operational systems may use additional scores for review, risk control, and publishability. They are not part of the public product methodology and are intentionally omitted from product endpoints.