Forensic signals, built to fuse with your technicals
Market Scholar runs a forensic engine over every market narrative and a separate, firewalled technical lane over price and volume. We deliver both as structured feeds — so an AI agent, a quant book, or a surveillance desk can take an RSI signal and ask the question a chart can’t: is the story behind this move actually true?
Two orthogonal axes beat either one alone
Technicals tell you when price is stretched — oversold, breaking out, trending. The forensic layer tells you whether the narrative behind that price is structurally trustworthy. The two are independent by construction: the technical lane reads OHLCV only and never sees a verdict; the forensic engine never sees the chart.
That independence is the point. Combining them removes a class of losers a technician can’t see and a forensic short-detector can’t rank — fewer setups that detonate on a manufactured or unsupported story.
- Measurement, not alpha. The forensic verdicts predict narrative structure, not returns — verdict edges collapse under beta/sector controls. We sell them as a risk filter and surveillance layer, not a signal to trade directly.
- One directional exception. WKS directional energy is the single signal that survives beta/sector/momentum controls — use it jointly, test it out of sample.
- Technical lanes are candidates. The playbook and momentum feeds carry decades of published evidence, but this dataset is dense only since early 2026 — one regime. Flags are screens, not validated live edges.
- Fair value is a guardrail. Stable but the least-calibrated subsystem — use it to cap how far you chase, never to size or pick direction.
Built for machines and the desks that run them
AI agents
Machine-readable second opinion for autonomous trading and research agents.
Your agent already generates technical setups. Call the forensic feeds as a pre-trade gate: skip or shrink any long where drift or coordination is elevated; keep full conviction where verification is high and coordination is low.
Quant funds
An orthogonal, OHLCV-independent narrative-structure factor for your risk overlay.
Fuse the forensic feeds with your own RSI / Bollinger / momentum signals as a position filter — concentrate exposure, suppress false positives. WKS is the one feed offered as carrying residual directional information; test it jointly, out of sample.
Banks & compliance
Surveillance and research — coordinated-campaign and source-degradation detection, no return claim required.
Detect orchestrated narrative campaigns, slop/banned-wire swarms, and analyst-vs-filing divergence across your universe. Pair coordination (the manipulation read) with the trading-footprint read to flag names where an orchestrated story meets anomalous tape.
The feeds
Every feed is a typed, point-in-time table. Read API feeds are directly queryable; Licensed feeds are provisioned per agreement.
Narrative forensics
The structural read on a story — how it verifies against the filings, who is pushing it, and whether it is coordinated.
The flagship vector — ~194 scored columns per company per run, plus a classified verdict. The single feed most consumers start from.
Verifies narrative claims against the latest 10-K/10-Q — pulls SEC revenue/EPS/margins and scores the Verification Match Score (text vs table).
Cross-source timing and phrasing analysis — clustered sources in a window, repeated phrasing, anonymous sourcing — scored to a coordination class.
A per-publication reliability score earned forensically from each outlet’s material-discrepancy-free rate on adjudicable claims (Wilson lower bound).
Per-analyst-by-ticker call track record — counts, bullish hit rate, realized 5/10-day post-call returns, consistency, and an edge score.
Checks a narrative’s underlying evidence premise (mechanism, not instance) against an embedded scholarly corpus — stance with a verbatim-quote gate.
Directed money-flow edges between AI-capex entities (investments, equity stakes, compute purchases) and the reciprocal round-trips they form.
Directional & valuation
The signals with a directional or valuation read — including the one forensic signal that survives beta, sector and momentum controls.
A signed accumulation/distribution energy signal (Walsh/Wyckoff) that replaced the deprecated sentiment channel — the strongest forensic signal in testing.
A pure-fundamentals fair value (blended FCF/EPS multiple, AI-tier adjusted) and the price premium/divergence against it, with a valuation verdict.
Per macro-theme roll-up of trap pressure — trap count/intensity, average coordination/drift/risk, and a contagion level with the worst name.
Per-company sector context — assigned sector/subsector, market size (2024 vs 2030) and CAGR, and narrative SWOT.
Technical & quant signals
Classic price-and-volume playbooks and momentum — the traditional technicals (RSI, Williams %R, Bollinger, ATR, relative strength) delivered natively, ready to fuse with the forensic feeds.
Eight research-backed short-term playbooks (Connors RSI-2, Williams %R, Bollinger, IBS, Donchian, Darvas, 52-week-high, Lehmann reversal) with ATR brackets ≥ 2.0 reward:risk.
Cross-sectional momentum and relative strength — RS vs SPY and sector, trend, breakout proximity, acceleration, volume confirmation — the “rising winners” axis.
Daily price/volume plus fundamentals — implied and 5-year-median P/E, trailing EPS, TTM revenue and YoY growth, average volume.
Narrative lifecycle
The full ledger of every observed narrative instance, its lineage, and its forward path.
A ~670K-row ledger where each “dot” is an observed narrative with genesis/chain lineage, cycle phase, direction, an embedding, and forward returns.
RSI lives in the feed, next to the verdict
You don’t bolt the technicals on afterward — the classic indicators ship as real, populated columns alongside the forensic scores. A sample of what’s live today:
Forensic + technical, worked through
RSI(2)-oversold, gated by verification
A name triggers Connors RSI(2) < 10 above its 200-day MA — a textbook mean-reversion long. Verification is high, drift is low, coordination is normal: take it at full conviction. The same trigger on a second name shows drift = 100 and a coordinated narrative — skip it. The technical signal is identical; the forensic overlay separates the trustworthy bounce from the trap.
Williams %R deep-oversold, suppressed by coordination
A stock hits Williams %R < −90 (deep oversold). The coordination feed shows a wire-swarm and a likely-coordinated class in the same window, while the trading-footprint read still looks normal — the “orchestrated narrative, no obvious crime” tell. The fused decision is to stand aside or shrink size; the bounce can fail on the next coordinated push.
Breakout, capped by fair value
A name breaks its 20-day high on rising volume (Donchian) and WKS agrees — entry confirmed. But fair-value divergence shows it already trades far above its fundamental anchor. Take the breakout, but use FVD as a chase-guard to cap the target rather than riding it open-ended. FVD never picks direction or size — it only limits how far you chase.
Momentum long, filtered by WKS agreement
The relative-strength lane ranks a name a top rising-winner vs SPY and its sector. Before committing, check WKS: keep the long where directional energy is still building; fade or pass where it is exhausting. Because WKS is the one forensic signal that survives beta/sector/momentum controls, this agreement filter is the most defensible directional fusion in the stack.
How you consume it
- Typed, point-in-time tables. Every feed is a Postgres relation keyed by ticker and snapshot date — no look-ahead, query the exact state on any date.
- Read API or direct SQL. Read-API feeds are queryable over HTTPS (REST/JSON); licensed feeds are provisioned with scoped credentials or a data share.
- Refreshed on the pipeline cadence. Forensic feeds update roughly six times a day; the technical lanes run pre-market and intraday.
- Licensed as derived data. Scores and structure, not raw third-party content — built for compliance review.
Wire it into your stack
Tell us which feeds and how you consume them — we’ll set up a scoped sandbox.
Derived research data, not investment advice. The forensic feeds are validated as predictive of narrative structure, not returns; the technical lanes are candidate screens calibrated over a single market regime. Backtest any overlay before relying on it.