Score Catalog¶
Canonical list of deterministic scores returned by CLI, Python, HTTP, and MCP.
Scoring version
This page describes deterministic_scoring_v2, the default for CLI, HTTP, MCP, and score_filing_html. For legacy v1 differences, see Versioning and Reproducibility and the v1 section in Aggregation.
Headline score¶
Field |
Range |
Meaning |
|---|---|---|
|
0–100 |
Weighted mean of present headline components (higher = more concern) |
|
0–1 |
Fraction of headline components computed (non-null) |
|
0–1 |
Blend of extraction quality, diff confidence, and coverage |
|
list |
Component names that could not be computed |
Weights are defined in COMPONENT_WEIGHTS (disclosure_alpha.scoring_types).
Headline components (nine)¶
Nine deterministic components feed the headline overall_disclosure_risk_score (weights in methodology/aggregation). specificity_quality_score is also returned but excluded from headline weights — see the score scale include for its inversion.
Plain English |
JSON field |
Primary section(s) |
|---|---|---|
Risk-factor tone & volatility |
|
Item 1A |
Year-over-year disclosure change |
|
Item 1A, MD&A |
MD&A uncertainty & demand stress |
|
Item 7 (10-K) / Item 2 (10-Q) |
Legal & regulatory risk language |
|
Item 1A (+ flags) |
Liquidity & covenant stress |
|
MD&A (+ flags) |
Boilerplate & vague risk language |
|
Item 1A |
Internal controls weakness signals |
|
Controls disclosure + Item 1A |
Material event severity (diff-only) |
|
Item 1A |
Cross-section negative tone |
|
Item 1A + MD&A |
Component |
Weight |
|---|---|
|
0.20 |
|
0.15 |
|
0.15 |
|
0.10 |
|
0.10 |
|
0.10 |
|
0.05 |
|
0.05 |
|
0.05 |
Blend formulas (v1): Aggregation.
v2 overrides (opt-in only)¶
These four components use different blends when you call score_deterministic_v2(). Headline weights and component names are unchanged; only the formulas and provenance differ.
Component |
v1 summary |
v2 summary |
|---|---|---|
|
|
Calibrated percentile ranks for tone ratios + diff |
|
Item 1A litigious + legal delta + flag boost |
Multi-section litigious evidence, legal delta, flag evidence (no +15 boost) |
|
MD&A constraining + liquidity density + flag boost |
MD&A evidence + Item 1A fallback + flag evidence |
|
Controls diff + Item 1A constraining + flag boost |
Controls-section diff + constraining + serious IC flag evidence |
v2 can emit non-null scores from flags alone when required tone metrics are absent. v1 often leaves those components null without Item 1A or MD&A text.
v2-only supplementary fields (opt-in)¶
Not in the v1 headline blend; populated by score_deterministic_v2() when evidence exists:
Field |
Meaning |
|---|---|
|
Specificity / inverse boilerplate (current filing) |
|
Tone, legal, liquidity, controls without YoY change |
|
Change, event severity, language deltas |
|
Incident flags (Item 1.05 / cyber sections) |
|
8-K event materiality proxies |
cybersecurity_incident_risk_score can be non-null on 10-K when incident flags fire in Item 1A (see Examples Gallery). event_materiality_score is typically non-null only when 8-K event sections are extracted — it stays null in the committed 10-K examples.
overall_disclosure_risk_score weights are unchanged in v1 and v2.
Supplementary component¶
Field |
In headline? |
Notes |
|---|---|---|
|
No |
Higher = better specificity (inverse of most risk scores) |
Aggregates¶
Field |
Meaning |
|---|---|
|
|
|
Mirrors |
Conditional nulls¶
Components are null (never zero) when required inputs are missing:
No Item 1A → Item 1A-derived components null
No MD&A → MD&A-derived components null
No prior filing →
disclosure_change_score,event_severity_score, and related diffs null
Full-coverage example with prior: Examples Gallery.