Veridion Markets/Backtest evidence/Track record
Methodology v2.13

The Veridion Score is a calibrated, coverage-aware composite.

The score maps six factor categories to a 0-100 scale. When a factor lacks usable data, the composite reweights across the factors that remain. Missing data is not filled with zero. Exact weights and normalization parameters stay server-side. Score calibration is anchored to a stable broad reference universe, so a ticker is not reranked just because the screener universe changes.

v2.13 tightened score integrity: earnings moved from coarse buckets to a continuous surprise curve, sparse analyst and news samples now shrink toward neutral, and high confidence is reserved for names with complete real factor coverage. Earlier v2.12 snapshots remain frozen as their own historical record.

Veridion Score methodology changelog

v2.13 integrity update

Shipped July 10, 2026

Before we turned marketing on, we audited our own flagship: the Veridion Score. We found two defects. Both are fixed in v2.13.

What we found

  1. 1. Earnings factor was degenerate. Under v2.12, the earnings factor collapsed to two coarse values dominated by a neutral default. It looked informative without producing a continuous distribution.
  2. 2. Confidence label was not backed by data. High-confidence labels could appear despite missing factor inputs. The label described model certainty without requiring complete underlying evidence.

What changed

  • Earnings now follows a continuous surprise curve.
  • Sparse earnings history is marked unavailable instead of receiving a favorable default.
  • Thin analyst and sentiment samples shrink toward neutral.
  • High confidence requires measured data across all six factors.
  • v2.12 history is frozen. We are not retroactively rescoring the past.
  • The model-integrity gate blocks deploys when a factor degenerates or confidence is not backed by coverage.
STRONG
75-100 + high confidence
High-conviction composite with broad factor coverage behind the score.
BUY BIAS
60-74, or 75+ with thinner confidence
Constructive read, including high scores that do not clear the STRONG confidence gate.
NEUTRAL
45-59
Balanced evidence or incomplete signal strength across the available factor set.
SELL BIAS
30-44
Unfavorable composite where risk evidence outweighs support.
AVOID
0-29
Adverse public band under the current methodology.
01 · Sentiment
Definition

News insight balance for the ticker across the recent article window.

Data source

Market news insights, filtered to ticker-level sentiment in the recent article window.

Forward-return justification

Forward-return review is reported on the public backtest page once band coverage clears the sample-size rule. View backtest evidence or review the public track record

Known limitation

News tagging can lag breaking events and can miss articles that do not carry ticker-level insight tags.

02 · Valuation
Definition

Market-universe valuation discipline from available company fundamentals and comparable multiples.

Data source

Financial statements, current quote data, and the covered-universe peer set.

Forward-return justification

Backtest evidence compares score bands across the same broad reference universe. View backtest evidence or review the public track record

Known limitation

Sparse statements can reduce usable metric coverage. Missing valuation fundamentals are excluded from the available subset.

03 · Hype
Definition

Attention velocity, abnormal volume, and breakout pressure measured from real market and news signals.

Data source

Daily market bars and ticker-level article counts, with financial safeguards from company statements.

Forward-return justification

Hype is validated inside the composite rather than published as a stand-alone forecast. View backtest evidence or review the public track record

Known limitation

Attention can fade quickly, and deeply unprofitable companies have capped upside from this factor alone.

04 · Analyst rating
Definition

Current analyst rating mix transformed onto a 0-100 scale, with thin coverage pulled toward neutral.

Data source

Analyst rating tape covering buy, hold, and sell-side rating counts.

Forward-return justification

Forward-return tables are linked by score band and include sample size plus standard error. View backtest evidence or review the public track record

Known limitation

Coverage differs by ticker. Sparse analyst counts are treated as lower-confidence evidence, not a full-strength signal.

05 · Momentum
Definition

Recent price return and moving-average structure from daily bars.

Data source

Daily aggregate market bars, using the available historical window when present.

Forward-return justification

The backtest page shows realized returns for the composite bands that include this factor when data exists. View backtest evidence or review the public track record

Known limitation

New listings and sparse bar histories can reduce the factor's usable coverage.

06 · Earnings
Definition

Recent earnings surprise history transformed into a continuous score when enough estimate-backed reports exist.

Data source

Earnings reports with actual and estimate fields when both are present.

Forward-return justification

Band-level forward returns are shown with null returns dropped rather than filled. View backtest evidence or review the public track record

Known limitation

Single-report or estimate-sparse histories are withheld from the factor until coverage clears the evidence floor.

Known limitations
  • Coverage gaps are explicit. A missing factor is dropped from the available subset rather than filled with zero.
  • Confidence rises with breadth of available factor coverage. High confidence requires all six scored company factors to have usable data.
  • Sparse samples are shrunk toward neutral rather than allowed to create a high-conviction score by themselves.
  • The highest public band requires both score strength and broad data coverage.
  • Valuation metrics can lag fundamental filings and can be unavailable for new or thinly covered issuers.
  • Newer tickers can have limited bar history, limited filings, or no forward-return window yet.
  • Backtest rows without complete 30d, 60d, and 90d returns are excluded from public band tables.
  • The score is descriptive research infrastructure. It is not financial advice.
Weekly Veridion brief

Rating changes, public disclosure activity, methodology notes, and product updates. One email per week. No advertising list resale.