About / Field Manual

Methods & Data

Implemented factor definitions, signal contracts, source freshness, limitations, and vocabulary

Overview

RankAlpha is an evidence and decision-support workspace for self-directed quantitative investors. It combines market context, cross-sectional factors, rule-based screens, event research, and monitoring. The frontend presents API and database outputs; it does not calculate a second set of grades in the browser.

Required grade inputs

Momentum + Value

Optional grade input

Sentiment

Default ranking order

Grade band, then symbol

Factor model cards

Purpose, implemented transformation, output, and limitation for each grade input.

Momentum

Purpose

Compare the strength and persistence of price trends across the covered universe.

Inputs

Canonical adjusted-close history over 200, 90, 50, and 30 trading-day windows.

Transform

Log-price regression slopes use configured base weights multiplied by a capped squared t-stat factor. The blended slope is ranked both raw and after sector-median residualization; the published Momentum score combines those percentiles using configured residual/raw weights (70/30 defaults).

Output

A daily 1–100 Momentum score.

Limitation

A high score describes relative trend strength; it is not a return forecast.

Value

Purpose

Compare relative cheapness using available fundamental valuation measures.

Inputs

EV/EBITDA, EV/FCF, P/FCF, P/S, P/B, and P/E fallbacks, plus dividend yield where a sector override uses it.

Transform

Cheapness percentiles blend peer-group and market context, apply Financials and Real Estate overrides, dynamically reweight available metrics, and shrink low-coverage results toward neutral.

Output

A daily 0–100 Value score.

Limitation

Coverage depends on vendor fundamentals, peer-group sample size, metric recency, and the applicable sector override.

Sentiment

Purpose

Add current analysis context when the persisted sentiment proxy is available.

Inputs

The latest stored AI scenarios and the price data used by the database sentiment-proxy view.

Transform

Scenario-weighted expected return is robustly scaled cross-sectionally in the database view.

Output

An optional 0–100 Sentiment score.

Limitation

Missing Sentiment does not penalize a stock; available factor weights are renormalized.

Composite & grade

Purpose

Combine required Momentum and Value evidence with optional Sentiment.

Inputs

Latest eligible factor scores and the effective model weights: the latest persisted version when available, otherwise the supported 0.40/0.40/0.20 defaults.

Transform

Weights are renormalized over available inputs, then the composite is ranked across eligible stocks.

Output

A 1–100 composite percentile mapped to grade bands A–F.

Limitation

The grade is a relative cross-sectional band, not a recommendation or expected return.

Signals and screeners

These Ideas workflows answer different questions and retain distinct ordering contracts.

Rankings

Compare the factor composite across the eligible universe. Default rows use grade bands A–F and alphabetical symbols within each band.

Open Rankings

Screener

Filter stored rule hits and consensus evidence by style, source, bias, and coverage.

Open Screener

Opportunities

Review the existing ranked multi-evidence candidate contract with thesis and invalidation context.

Open Opportunities

Earnings Watch

Monitor the Calendar subset explicitly classified as earnings surprise candidates.

Open Earnings Watch

Data and freshness

Freshness is scoped to a subject. Current, delayed, stale, and unavailable states describe the named module, not the entire application. Pipeline cadence is an expectation, not proof that a particular run succeeded.

Module	Source boundary	Expected cadence	Displayed evidence
Prices & technicals	Daily market observations	Nightly after source data is available	Price/technical timestamp at point of use
Grades & factor scores	Cross-sectional scoring pipeline	Nightly when the scorer completes	Score date plus scorer status
Model weights	Hybrid information-coefficient and risk-parity process	Monthly unless explicitly recomputed	Weight version/effective date when available
AI research	Stored memo and evidence pipeline	Varies by research trigger and coverage tier	Analysis date, source, and quality status

Scenarios

Bull, base, and bear cards show memo-authored probability and target values with the analysis date and source. The current persistence contract does not store an exact AI reference price, so the product does not derive a scenario return from the current quote. If a future contract persists an exact reference price and date, the return is computed as (target − reference) ÷ reference.

Scenario probabilities are research weights, not guarantees, trading instructions, or calibrated odds unless a separately published validation says so.

Validation and limitations

Implemented controls

Canonical price schema and deterministic latest-score tie-breaks.
Momentum and Value coverage gates before a grade is eligible.
Dynamic factor reweighting when optional Sentiment is missing.
Point-of-use dates and module status where the source exposes them.

Unavailable for publication

A product-wide, independently reviewed live-performance history and calibrated scenario-probability study are not published in the current repository. RankAlpha does not imply those results. Some research modules expose their own scoped validation and must be read within that module’s stated sample and protocol.

Change log

Public model change history: unavailable as a consolidated publication.

The controlling implementation remains versioned in database migrations, scorer code, and repository documentation. This page will not invent release dates or backfill performance claims.

Glossary

As of / score date: The effective date of the named dataset. It does not describe every module on the page.
Composite percentile: A stock’s relative position among eligible composites for the same scoring snapshot.
Confidence: A label for evidence coverage or internal consistency; it is not a probability of gain.
Grade band: A–F bucket derived from the composite percentile.
Market session date: The Eastern market date used to keep a Today payload internally consistent.
Scenario probability: The memo-authored weight attached to a bull, base, or bear scenario.
Setup hit: A stored observation that a named screener rule fired during its defined window.

Privacy, security posture, legal terms, and investment disclaimers live in Trust & Legal.