SEC filing scoring FAQ

No. Disclosure Alpha uses a deterministic pipeline: section extraction, text metrics, boolean risk flags, and year-over-year diffs. Scores are fully reproducible given the same input HTML and version strings. No LLM is required for scoring.

10-K and 10-Q are supported via EDGAR ticker routes (CLI, Python SDK, HTTP API, MCP Analyst) or local --html scoring. 8-K event text is supported via local --html or MCP Builder only — not via EDGAR or HTTP ticker routes.

Nine headline-weighted component scores (ten computed overall) combine into overall_disclosure_risk_score (0–100) using deterministic_scoring_v2. Components cover risk-factor language, boilerplate, MD&A uncertainty, liquidity stress, disclosure change, and controls signals. Weights renormalize when components are missing.

deterministic_scoring_v2 is the default scoring model version string returned in every score response. It defines component weights and aggregation rules. Same version strings and input text produce identical scores across CLI, Python SDK, HTTP API, and MCP.

No. Disclosure Alpha output is for research and integration testing. It is not investment advice, not a trading signal, and not return prediction. Read the underlying SEC filings before making decisions.

Install with pip install "disclosure-alpha", then run disclosure-alpha score --html path/to/filing.html --form 10-K. No SEC User-Agent or network required. For year-over-year comparison, add --prior-html with a prior filing.

Use CLI or Python SDK for scripts, notebooks, and offline local HTML. Use HTTP API for services, dashboards, and batch panel screening (up to 25 tickers). Use MCP Analyst for agent workflows with ticker + fiscal year; use MCP Builder for raw HTML agent pipelines including 8-K.

Automated validation on 478 S&P 500 FY2025 Item 1A extractions (universe n=503), post quality filters. Specificity construct validity: Spearman ρ ≈ 0.87 vs NER entity density (n=478). Boilerplate construct validity: Spearman ρ ≈ 0.92 vs ls_boilerplate_word_ratio on boilerplate_combined_ratio (text_metrics_v4, n=478); phrase-only v3 was ≈0.74. Post-filing volatility association Q5/Q1 ≈ 1.15 on n=435 cohort (90-day realized vol) — descriptive association only — not return prediction.

8-K scoring is available via local --html or MCP Builder with raw HTML only. It is not supported on EDGAR ticker routes or HTTP API ticker endpoints. Agent builders should use disclosure-alpha-mcp-builder for 8-K event text pipelines.

Every score response includes artifact version strings (e.g. section_extractor_v1, text_metrics_v4, deterministic_scoring_v2, built_in_dictionaries_v3). Pin package version from PyPI and record input HTML plus version strings for reproduction. Corpus layout and scripts: data/validation/ on GitHub.

score_coverage_ratio reflects how many expected filing sections were successfully extracted and scored. Missing sections reduce coverage and confidence. Inspect extraction_warnings and section text when coverage is low before trusting scores at scale.

For installation errors, SEC EDGAR setup, and common CLI issues, see the ReadTheDocs FAQ. For methodology and product positioning, see /methodology and /why-deterministic. RTD troubleshooting: https://disclosure-alpha.readthedocs.io/en/latest/getting-started/faq.html