EventHorizonIQ  /  Independent research

Independent research for prediction-market decisions.

See where EHIQ disagrees with market expectations, understand the evidence behind each call, and check how past calls turned out. Decide for yourself whether the research changes your view. Each published call carries its open date, resolved calls are graded in public, and the misses are kept as loudly as the hits.

0 / 100
Structural Risk Index (SRI) · LOW
Structure is healthy.

Calls dated before the outcome. Hits and misses kept in the public record.

Recent calls landed: Iowa Senate D +31pp vs Kalshi (Jun 3, HIT) · California Governor +15pp vs Polymarket (Jun 3, HIT) · LA Mayoral 2-slot package HIT (Jun 3) · BCRED gating (Pebbles III, Jun 4, published two hours before the gate fired) · Iceman exit-architecture series — 5 layers →
Today's read  /  RATE DNA  /  2026-10-02
The jobs report was a growth scare, not an all-clear.
+29K vs ~90K. Real yields −2.9bp, breakevens −1.5bp, cuts priced at the front, fiscal pressure intact at the back. Rate relief from growth damage is the bad flip — it deepens the squeeze on beaten-down names; it doesn't end it.
Read the decomposition →

Active calls · 37

EHIQ probability vs market, before each resolves

All calls →

Firing now · 6

Sensors above STABLE right now

All sensors →
10 days
before the Feb 2026 BTC drawdown
RECONSTRUCTED · Jan 26 → Feb 5, 2026 · $87,974 → -28.3%
60 days
before the 2023-03-10 SVB failure
BACKTEST · triggered before 7 of 7 US bank failures, 2008-2024
$1.40 → $95.69
Carvana, Jan 2023 signal print
RECONSTRUCTED · 68x, CVNA US Equity · no signal was published at the time
1 quarter
before the 2025 Apple AI drawdown
BACKTEST · TonalityIQ patterns triggered one quarter earlier
The Tell · TonalityIQ CEO series

Complete earnings-call corpora, read CEO by CEO: the language patterns, inflection quarters, and tells the numbers miss.

Read the series →
Ask EHIQ

Powered by local AI. Not investment advice.

Get the next call before the crowd.

EHIQ flagged the Musk–Altman split 9 days before it broke. Subscribe free for the methodology breakdown — plus an alert the moment a sensor flips or a tracked call resolves.

Free forever. Unsubscribe anytime. No spam.

Free channels

Structural Risk Index (SRI)

Structure, not sentiment.

LOW
0 / 100
LOWMODRISINGELEVSEVERE

Structure is healthy.

Last SEVERE: May 2021 (81/100), BTC fell 53%. Jan 2026: ELEVATED (55/100), BTC fell 33% in 10 days.

14 sensors elevatedContinuously-audited sensor baseFull index →Methodology →

Validated against BTC 30-day forward drawdowns (IC -0.31). BTC leads the cycle: when structure breaks, everything follows.

120 Stable
9 Rising
13 Elevated
142 sensors active

Stress Signals — 13 sensors elevated

MacroMEDIUM

Correlation to One

--
SEVERE
4990
11d stale

The correlation-regime trigger for the rate-pressure family and the board (Kawa 2022 framework, spec stamped 2026-10-03): when cross-asset correlations go to one, stop tilting and control risk. CORRELATIONS-COUPLING carries risk_control_override=True for the board/CONCERT — a risk-control override, never a directional call. Measure (v2.0, frozen): sign-adjusted 60-day mean correlation across the estate's core legs (risk-risk +1; risk-yield -1; risk-credit -1; yield-credit +1), with unsigned |corr| printed as the shock-coupling corroborator. v1.x unsigned measure falsified for 2022 and replaced — declared in the build memo.

MacroHIGH

Growth Valuation

--
SEVERE
4990
8d stale

Tech-sector earnings yield vs real 10Y Treasury yield. Detects when growth multiples are extreme relative to the prevailing rate environment.

MacroMEDIUM

Sector Rotation

--
SEVERE
4990
8d stale

Capital rotation from tech/growth into defensives: XLK vs XLV/XLF/XLE/XLU sector ETFs. Tech underperforming defensives by >5% over 21d marks a rotation underway — the move that precedes or accompanies growth selloffs.

MacroMEDIUM

AI Capex Stress

--
ELEVATED
4990
8d stale

Credit-market stress on AI hyperscaler capex funding: ICE BofA US High Yield OAS (FRED BAMLH0A0HYM2) combined with QQQ drawdown from all-time high. When HY spreads widen, tech capex gets more expensive/risky — detects environments where the massive-AI-capex names (ORCL, MSFT, META, GOOGL) face higher financing costs into the build-out.

MacroMEDIUM

Commodities Ballast

--
ELEVATED
4990
11d stale

The 60/40 failure condition (Kawa 2022 framework, spec stamped 2026-10-03): when equities and bonds fall TOGETHER bonds are not hedging — and this sensor reads whether commodities are providing the replacement ballast. Regime on the 63-day SPX/TLT correlation and joint drawdown: BONDS-HEDGING (corr < 0), BOTH-LEGS-FAILING (corr >= 0 with both legs down), MIXED otherwise; inside BOTH-LEGS-FAILING, BCOMTR positive = BALLAST-PRESENT else BALLAST-ABSENT. Disjoint from kawa_rate_pressure by construction: kawa measures the credibility shock in inflation EXPECTATIONS, this measures the realized portfolio failure in cross-asset RETURNS (Jaccard 0.115). Ledger value = corr_63.

MacroMEDIUM

Compute Cost-and-Constraint Watch

--
ELEVATED
4990
8d stale

The price of what the AI buildout buys, and the physical brake on it — the Creativity Watch's third leg (spend = the CapEx Medusa; price and brake = here). Monthly composite of constraint legs: electric-power PPI (the regulated T&D 'who pays' leg), copper (transformer/equipment cost), warehouse-construction PPI (the genre data centers live in), and public frontier-model API price lists (inference cost deflation, via the litellm community price file's git history — timestamped by commit). GPU spot (vast.ai H100-class median $/hr + p90/p10 spread) rides as metadata, accumulating from 2026-10-03. LBNL interconnection queue: DECLARED ABSENT via automated path (403 to bots; parked manual-curate ask). States: CHEAPENING (deflation dominates) / STABLE / CONSTRAINED (the brake is binding). FALSIFIER RESULT, printed full-size: the frozen lead-lag bar vs the estate's own Medusa (BEST_CAPEX revisions) FAILED — rho_lead1 -0.43 vs rho_0 -0.48: the composite is COINCIDENT with the Medusa, not leading. It prints as a DESCRIPTIVE constraint gauge (DeVries negative control 0/18 false fires; the 2026 power-crunch window reproduced); the leading-instrument claim is null.

MacroMEDIUM

Global Risk Appetite

--
ELEVATED
4990
8d stale

Non-US equity momentum as a global risk-appetite proxy: KOSPI, DAX, Nikkei (yfinance). Mike L's finding: KOSPI 21d returns are predictive of BTC. Strong negative momentum abroad signals the broad risk-off that precedes or accompanies crypto deleveraging.

FundamentalMEDIUM

Market Quality Gauge

--
ELEVATED
4990
8d stale

The meta-instrument grading the register's market inputs: a call's market_prob_at_open is only as good as the market it came from. Per tracked event: DEEP (thick volume, tight spread, smooth action — read at face), THIN (low volume or wide spread — width caveat), WALLET-DRIVEN (deep volume but price jumps on small trades — manipulation shape), and CONFLICTED when two events price the SAME question >10pp apart: (raw) two tracked events genuinely apart — the state that invalidates a probability, with the legs' own grades (DEEP/THIN) carried alongside since the divergence group has no thin filter by design; (computed) a bands-implied approximation vs a binary, where part of the gap is measurement shape — a caution, not an invalidation (renamed from DIVERGENT 2026-10-04, Eric's stamp: divergence here means EHIQ-vs-market; this is markets diverging from each other).

MacroMEDIUM

NASDAQ Momentum

--
ELEVATED
4990
8d stale

US tech momentum via QQQ: drawdown from the 60d high plus 21d momentum. The primary equity top/bottom detector — catches broad tech selloffs (the Sep 2025 ORCL $328→$136 shape) that macro sensors alone miss.

MacroMEDIUM

Polymarket IQ

--
ELEVATED
4990
8d stale

Aggregated belief from Polymarket prediction markets on macro, geopolitical, and regulatory events. Tracks crowd-belief divergence from consensus.

Macrohigh

Real Rate Regime Sensor

--
ELEVATED
4990
8d stale

Real yields elevated — significant headwind for equities and crypto.

FundamentalMEDIUM

Reversion Speed

--
ELEVATED
4990
10d stale

The tape's mean-reversion speedometer (clock-2): the median half-life (days to >=50% retrace, cap 20d) of >=2-sigma single-name moves across the book's names + spec basket + index controls. FAST (trailing-30d median well below the 24-month band) = churn/dispersion regime — reversions machinery speeding up; SLOW (well above the band) = trending-stress regime — >=2-sigma moves do NOT revert; NORMAL inside the band; LOW-N (<5 qualifying events, declared, never a fake median). The capped share (moves that never reverted) is part of the state print, not a footnote. Fleet map: FAST -> ELEVATED, NORMAL -> NORMAL, SLOW -> STABLE, LOW-N -> NONE. Daily recompute of the trailing 30-day event window; staleness > 3d -> mock (refused at the ingest door).

MacroMEDIUM

Sovereign Credibility

--
ELEVATED
4990
9d stale

The ECB's TPI trigger logic, instrumented: is a sovereign spread move WARRANTED (fundamentals-paced fiscal repricing the ECB can ignore) or DISORDERLY (the move the Transmission Protection Instrument was built for)? Per sovereign, per day: 4 disorderly signatures (gap-widen, cross-sovereign contagion, bank-equity contagion, FX dislocation — auction failure DECLARED ABSENT for v1) drive a 3-state machine with hysteresis: CONTAINED (0) / WATCHING (exactly 1) / CONTAGION (>=2).

Macro Sensors

Crypto Sensors

Other Sensors

MacroMEDIUM

Correlation to One

--
SEVERE
4990
11d stale

The correlation-regime trigger for the rate-pressure family and the board (Kawa 2022 framework, spec stamped 2026-10-03): when cross-asset correlations go to one, stop tilting and control risk. CORRELATIONS-COUPLING carries risk_control_override=True for the board/CONCERT — a risk-control override, never a directional call. Measure (v2.0, frozen): sign-adjusted 60-day mean correlation across the estate's core legs (risk-risk +1; risk-yield -1; risk-credit -1; yield-credit +1), with unsigned |corr| printed as the shock-coupling corroborator. v1.x unsigned measure falsified for 2022 and replaced — declared in the build memo.

MacroHIGH

Growth Valuation

--
SEVERE
4990
8d stale

Tech-sector earnings yield vs real 10Y Treasury yield. Detects when growth multiples are extreme relative to the prevailing rate environment.

MacroMEDIUM

Sector Rotation

--
SEVERE
4990
8d stale

Capital rotation from tech/growth into defensives: XLK vs XLV/XLF/XLE/XLU sector ETFs. Tech underperforming defensives by >5% over 21d marks a rotation underway — the move that precedes or accompanies growth selloffs.

MacroMEDIUM

AI Capex Stress

--
ELEVATED
4990
8d stale

Credit-market stress on AI hyperscaler capex funding: ICE BofA US High Yield OAS (FRED BAMLH0A0HYM2) combined with QQQ drawdown from all-time high. When HY spreads widen, tech capex gets more expensive/risky — detects environments where the massive-AI-capex names (ORCL, MSFT, META, GOOGL) face higher financing costs into the build-out.

MacroMEDIUM

Commodities Ballast

--
ELEVATED
4990
11d stale

The 60/40 failure condition (Kawa 2022 framework, spec stamped 2026-10-03): when equities and bonds fall TOGETHER bonds are not hedging — and this sensor reads whether commodities are providing the replacement ballast. Regime on the 63-day SPX/TLT correlation and joint drawdown: BONDS-HEDGING (corr < 0), BOTH-LEGS-FAILING (corr >= 0 with both legs down), MIXED otherwise; inside BOTH-LEGS-FAILING, BCOMTR positive = BALLAST-PRESENT else BALLAST-ABSENT. Disjoint from kawa_rate_pressure by construction: kawa measures the credibility shock in inflation EXPECTATIONS, this measures the realized portfolio failure in cross-asset RETURNS (Jaccard 0.115). Ledger value = corr_63.

MacroMEDIUM

Compute Cost-and-Constraint Watch

--
ELEVATED
4990
8d stale

The price of what the AI buildout buys, and the physical brake on it — the Creativity Watch's third leg (spend = the CapEx Medusa; price and brake = here). Monthly composite of constraint legs: electric-power PPI (the regulated T&D 'who pays' leg), copper (transformer/equipment cost), warehouse-construction PPI (the genre data centers live in), and public frontier-model API price lists (inference cost deflation, via the litellm community price file's git history — timestamped by commit). GPU spot (vast.ai H100-class median $/hr + p90/p10 spread) rides as metadata, accumulating from 2026-10-03. LBNL interconnection queue: DECLARED ABSENT via automated path (403 to bots; parked manual-curate ask). States: CHEAPENING (deflation dominates) / STABLE / CONSTRAINED (the brake is binding). FALSIFIER RESULT, printed full-size: the frozen lead-lag bar vs the estate's own Medusa (BEST_CAPEX revisions) FAILED — rho_lead1 -0.43 vs rho_0 -0.48: the composite is COINCIDENT with the Medusa, not leading. It prints as a DESCRIPTIVE constraint gauge (DeVries negative control 0/18 false fires; the 2026 power-crunch window reproduced); the leading-instrument claim is null.

MacroMEDIUM

Global Risk Appetite

--
ELEVATED
4990
8d stale

Non-US equity momentum as a global risk-appetite proxy: KOSPI, DAX, Nikkei (yfinance). Mike L's finding: KOSPI 21d returns are predictive of BTC. Strong negative momentum abroad signals the broad risk-off that precedes or accompanies crypto deleveraging.

FundamentalMEDIUM

Market Quality Gauge

--
ELEVATED
4990
8d stale

The meta-instrument grading the register's market inputs: a call's market_prob_at_open is only as good as the market it came from. Per tracked event: DEEP (thick volume, tight spread, smooth action — read at face), THIN (low volume or wide spread — width caveat), WALLET-DRIVEN (deep volume but price jumps on small trades — manipulation shape), and CONFLICTED when two events price the SAME question >10pp apart: (raw) two tracked events genuinely apart — the state that invalidates a probability, with the legs' own grades (DEEP/THIN) carried alongside since the divergence group has no thin filter by design; (computed) a bands-implied approximation vs a binary, where part of the gap is measurement shape — a caution, not an invalidation (renamed from DIVERGENT 2026-10-04, Eric's stamp: divergence here means EHIQ-vs-market; this is markets diverging from each other).

MacroMEDIUM

NASDAQ Momentum

--
ELEVATED
4990
8d stale

US tech momentum via QQQ: drawdown from the 60d high plus 21d momentum. The primary equity top/bottom detector — catches broad tech selloffs (the Sep 2025 ORCL $328→$136 shape) that macro sensors alone miss.

MacroMEDIUM

Polymarket IQ

--
ELEVATED
4990
8d stale

Aggregated belief from Polymarket prediction markets on macro, geopolitical, and regulatory events. Tracks crowd-belief divergence from consensus.

FundamentalMEDIUM

Reversion Speed

--
ELEVATED
4990
10d stale

The tape's mean-reversion speedometer (clock-2): the median half-life (days to >=50% retrace, cap 20d) of >=2-sigma single-name moves across the book's names + spec basket + index controls. FAST (trailing-30d median well below the 24-month band) = churn/dispersion regime — reversions machinery speeding up; SLOW (well above the band) = trending-stress regime — >=2-sigma moves do NOT revert; NORMAL inside the band; LOW-N (<5 qualifying events, declared, never a fake median). The capped share (moves that never reverted) is part of the state print, not a footnote. Fleet map: FAST -> ELEVATED, NORMAL -> NORMAL, SLOW -> STABLE, LOW-N -> NONE. Daily recompute of the trailing 30-day event window; staleness > 3d -> mock (refused at the ingest door).

MacroMEDIUM

Sovereign Credibility

--
ELEVATED
4990
9d stale

The ECB's TPI trigger logic, instrumented: is a sovereign spread move WARRANTED (fundamentals-paced fiscal repricing the ECB can ignore) or DISORDERLY (the move the Transmission Protection Instrument was built for)? Per sovereign, per day: 4 disorderly signatures (gap-widen, cross-sovereign contagion, bank-equity contagion, FX dislocation — auction failure DECLARED ABSENT for v1) drive a 3-state machine with hysteresis: CONTAINED (0) / WATCHING (exactly 1) / CONTAGION (>=2).

Crypto

Eth Gas Fee Regime

--
RISING
4990
10d stale
CryptoHIGH

IV-RV Spread (Variance Risk Premium)

--
RISING
4990
9d stale

Deribit DVOL minus 30-day realized vol. When options overprice fear relative to realized conditions, subsequent BTC returns are positive. Correlation: 0.172 with 21d direction.

MacroMEDIUM

Mortgage Channel

--
RISING
4990
8d stale

Mortgage/housing transmission channel of financial conditions (Kawa/Odd Lots framework). Reads whether a rate move is actually transmitting through the 30Y mortgage rate into housing activity (starts + permits), not just whether rates are high. States: TRANSMITTING-TIGHT (tightening getting through), EASING (easing getting through), DRAG (easing but housing falling anyway — growth damage), ABSORBING (tightening, boom carrying on), STALLED / STALLED-HIGH (impulse present, channel clogged — lock-in), NON-RATE-MOVE, QUIET. Fleet state maps TRANSMITTING-TIGHT to RISING/SEVERE, STALLED-HIGH and DRAG to ELEVATED; the channel label rides in metadata with a read-through to the rate-sensitive consumer cohort (OPEN).

MacroHIGH

Treasury Yield Inversion

--
RISING
4990
8d stale

2Y-10Y Treasury spread monitor. Tracks the canonical recession-signal yield curve relationship. Distinct from yield-curve (slope dynamic) — measures inversion magnitude specifically.

Macro

CAD/USD Risk Barometer

--
NEUTRAL
4990
10d stale

Commodity currency risk-off detector (CAD, AUD, NOK vs oil)

Macro

Canada Macro Composite

--
NEUTRAL
4990
10d stale

Canadian economy leading indicator (CAD, TSX, oil, employment, housing, BoC)

Crypto

Eth Eip1559 Burn Rate

--
NEUTRAL
4990
10d stale
MacroMEDIUM

Affiliated-Insurance Credit Watch

--
STABLE
4990
16d stale

The PE-insurer private-credit backstop, read from the insurer side (ODD LOTS synthesis-1 standout #1: Granato/Drall 'Private Credit State Backstop' 2026-07-31 + Sheehan/Manchuck 2026-04-27). PE firms bought ~$750B of life insurers and park affiliated private credit on insurer balance sheets — the policyholders are the backstop for the PE firm's credit risk. The BDC scanner reads the funds; this watch reads the insurer side from SEC XBRL companyfacts, as-of by filing date. States: FUNDED / STRETCHED / BACKSTOP-AT-RISK over three legs — affiliated-complex concentration (APO/KKR/CRBG/BNT vs a 6-name non-PE control cohort), AOCI duration strain (-AOCI/equity change at APO+CRBG), and affiliated PIK share (MFIC/FSK/BXSL XBRL instance fields; MFIC = Apollo's filer post-2022 AINV merger). Fleet maps FUNDED->STABLE, STRETCHED->RISING, BACKSTOP-AT-RISK->ELEVATED (SEVERE at score>=5).

MacroHIGH

Dollar-Hedging Character Leg

--
STABLE
4990
11d stale

The cross-currency basis as the price of dollar insurance — a flow-based credibility vote on the dollar system that domestic rate instruments cannot see. RATE DNA credibility family: kawa_rate_pressure reads credibility in inflation expectations, sovereign_credibility in EU spreads, commodities_ballast in cross-asset returns; this leg reads it in global dollar-funding flows (BIS Shin 2025-10-23; Torman 2026-03-26). Production series is the derived 3M CIP basis from FX forwards + money-market rates (the vendor's direct quotes died 2021-12-31), validated against them over 2019-2021 (EUR corr 0.98, JPY corr 0.96). States: HEDGED / COMPLACENT / STRESSED / SCRAMBLING from the worst-of EUR/JPY/GBP stress z.

MacroHIGH

Front-End Plumbing (Reserve Regime)

--
NORMAL
4990
8d stale

The reserve-regime / front-end plumbing leg (Joe Abate framework, Odd Lots 2025-10-03): banks stopped trading fed funds when reserves turned abundant, pinning FF-IORB at ~-7bp for four years; money-market spreads are the honest barometer of whether the Fed's liquidity management is working. States ABUNDANT / AMPLE / SCARCE-WATCH / SCARCE from EFFR-IORB and SOFR-IORB spreads (persistence-gated for SCARCE), TGCR-SOFR tiering, RRP buffer level and drain, and WALCL balance-sheet direction. One sensor filling two declared gaps: RATE DNA's declared-UNAVAILABLE front-end (the compositor now prints the plumbing regime as its front-end clause) and dealer-capacity's declared-ABSENT repo leg (carried as repo-leg metadata, never a state driver there). v1.0.2 with two declared falsifier-driven corrections (flood-floor EFFR pin level; RRP-drain gate on WALCL direction), thresholds re-frozen in the docstring. Episode-verified: Sept-2019 repo spike SCARCE (SOFR-IORB +315bp), 2020 flood ABUNDANT, 2023-25 pinned era median EFFR-IORB -7.0bp (Abate's claim verified exactly), quiet windows clean. Information-add CONFIRMED vs kawa_rate_pressure (rho +0.04) and dealer_capacity (rho +0.16).

MacroMEDIUM

Issuance Calendar

--
NORMAL
4990
8d stale

Per-month issuance-calendar state over the Creativity Watch's hyperscaler/AI-infra name set (ORCL, META, GOOGL, AMZN, MSFT, NVDA + CRWV metadata-only). Reads whether the USD bond-issuance channel is flowing: OPEN (flowing), FRONT-LOADED (supply pulled forward — trail3/median >= 1.75), PAUSED (zero-USD month without crack pre-conditions), AIR-POCKET (zero issuance AND no front-load relief AND forward maturities exceeding demonstrated placement capacity AND credit widening — the crack pre-condition). EDGAR-primary (Bloomberg bond-level declared ABSENT after entitlement probes); ORCL CDS and CDX IG supply the stress legs; NON-USD deals ride as a native-currency footnote (the routing-around signature).

MacroMEDIUM

MOVE Index

--
STABLE
4990
8d stale

The ICE BofA MOVE Index — implied volatility of US Treasury options, the bond market's VIX. When MOVE spikes, rates uncertainty cascades into credit, equities, and crypto. Oct 2023 hit ~160 on the term-premium repricing.

MacroHIGH

Rate DNA

--
STABLE
4990
10d stale

Cross-asset long-end rate pressure per the Luke Kawa framework (Odd Lots): yields are set by the growth impulse transmitted through REAL yields, with credibility shocks as a separate mode. Speed, not level. Two modes: GROWTH-LED (default: z of real-10Y 4-week change + growth leg copper/SPXEW/SX5E + SPX leg, breakeven anchor, BTC-dominance and earnings-faith legs at w=0.5) and CREDIBILITY (inflation-swap z outside the 0.5 band: real-yield z plus clipped swap z, no equity leg). v5.0.0 stamped 2026-10-01 with frozen episode bars (2022 SEVERE+TIGHTENING >= 40%, zero EASING at the Oct-2022 peak) and a beaten naive baseline every year 2023-2026. This is the engine behind the public EHIQ RATE DNA print.

equityMEDIUM

SPEC Composite

--
NORMAL
4990
9d stale

The broad retail speculative-flow composite: composite z of four retail-speculative legs over the spec basket — (1) BJZZ+QMP small-trade imbalance (mean of per-name window z's), (2) <=7-DTE OTM call aggressive-buy notional (rolling-60 z), (3) basket 20d mean return (rolling-60 z), (4) basket pairwise co-movement (rolling-60 z). Composite = mean of the AVAILABLE legs (declared-absent, never zero-filled; n_legs printed on every read; zero available legs = NONE/LOW-N, never a fake composite). Raw states COLD / COOL / WARM / SPECULATING / MANIA on the composite z mapped STABLE / NORMAL / RISING / ELEVATED / SEVERE. Reads regimes, not names — the single-name companion is spec-singlename (companion, never pooled; the composite is structurally blind to single-name squeezes by construction, proven in the v1 NULL memo).

MacroHIGH

Strategic-Metals War Sensor (Tungsten-Primary)

--
NORMAL
4990
11d stale

War anticipation measured in metal demand, not headlines: military procurement shows up in tungsten APT before conflict (armor-piercing projectiles, cluster munitions; China-dominated supply; minimal civilian-cycle exposure — Fickling 2026-08-04). Tungsten-primary states off L_t = max(robust-z of 63d trend, 21d acceleration) on the Europe APT Rotterdam series: CALM / PROCURING (>= +1.5) / MOBILIZING (>= +2.0). Copper and SOX print as a breadth qualifier (metadata, never state-driving); the Europe/China APT wedge prints the bulcanization premium. The estate's geopolitical-risk sensor reads the GPR news layer; this sensor reads the procurement layer — disjoint inputs (Jaccard 0.270 over the 2026 overlap).

MacroMEDIUM

Wage Leg

--
LOW
4990
8d stale

The realized-wage channel of the rate-character family: what workers are actually paid (AHE y/y level + slope) vs what the breakeven market expects. Separates a DISINFLATIONARY SLOWDOWN (wages cooling, payrolls holding — the Fed can stop) from a STAGFLATIONARY SLOWDOWN (wages falling because the labor market is breaking). Raw states WAGES-ACCELERATING / STABLE / SOFTENING with a goods/services decomposition flag; the two-sided classifier pairs wage softening with the absolute payroll stall bar (3-month change < 100K — a relative-to-trend bar was tested and rejected for misfiring in every healthy deceleration). Fleet map: ACCELERATING -> ELEVATED, STABLE -> NORMAL, SOFTENING -> LOW. Value = AHE y/y.

alternative

App Store Rank: NXDR

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

iOS App Store rating count and rank movement tracker

alternative

App Store Rank: OPEN

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

iOS App Store rating count and rank movement tracker

alternative

App Store Rank: PTON

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

iOS App Store rating count and rank movement tracker

alternative

App Store Rank: ROOT

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

iOS App Store rating count and rank movement tracker

alternative

App Store Rank: UPST

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

iOS App Store rating count and rank movement tracker

CryptoHIGH

BTC Capitulation Composite Sensor

WITHHELD — mock data published as real: runner builds them with no data provider (eric-stamp ML 2026-10-16)
--
UNMEASURED
4994
28d stale

Composite gauge of 6 capitulation components for BTC market structure.

composite

BTC Defensive v0 (SHADOW)

--
UNKNOWN
4990
37d stale

Shadow candidate, stamped 2026-09-26. Validated features only, sign-corrected contrarians; DEFENSIVE when >=2 net stress signals fire. Writes a daily risk state; ALLOCATES NOTHING. The claim is the live record, graded independently by Adam's harness. Pre-registered scoring rule frozen in scripts/btc_defensive_v0_shadow.py — edits require a new sensor id.

composite

BTC Defensive v1 (SHADOW)

--
UNKNOWN
4990
37d stale

Regime-conditional increment, PREREG stamped 2026-09-26. v0 stress state gated by close-vs-SMA200: RISK_ON / BEAR_QUIET / STRESS_IN_RALLY / DEFENSIVE. ALLOCATES NOTHING. Adam-read precondition released by Eric 2026-09-26; price precondition met via frozen import (633 rows, sha256 df5b06c5). Rule frozen in scripts/btc_defensive_v1_shadow.py — edits require a new sensor id.

Fundamental

CEO Language Decay: BETR

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Tracks hedging vs confidence language in earnings transcripts

Fundamental

CEO Language Decay: BITF

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Tracks hedging vs confidence language in earnings transcripts

Fundamental

CEO Language Decay: CIFR

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Tracks hedging vs confidence language in earnings transcripts

Fundamental

CEO Language Decay: IREN

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Tracks hedging vs confidence language in earnings transcripts

Fundamental

CEO Language Decay: NXDR

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Tracks hedging vs confidence language in earnings transcripts

Fundamental

CEO Language Decay: OPEN

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Tracks hedging vs confidence language in earnings transcripts

Fundamental

CEO Language Decay: PTON

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Tracks hedging vs confidence language in earnings transcripts

Fundamental

CEO Language Decay: ROOT

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Tracks hedging vs confidence language in earnings transcripts

Fundamental

CEO Language Decay: SANA

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Tracks hedging vs confidence language in earnings transcripts

Fundamental

CEO Language Decay: UPST

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Tracks hedging vs confidence language in earnings transcripts

Macromedium

Canada Housing Market

--
DECLINING
4990
10d stale

Tracks Canadian housing from Statistics Canada: the new housing price index year-over-year change, with national, Toronto and Vancouver housing starts alongside.

composite

Capitulation Convergence

--
NONE
4990
10d stale
Macro

Corporate Stress Index

--
CALM
4991
11d stale

Corporate stress indicator derived from real-time search behavior patterns.

Macro0.5

Cross-Asset Spillover

WITHHELD — reads the mock BTC gauge (btc-capitulation-gauge) as its crypto leg (eric-stamp ML 2026-10-16)
--
UNMEASURED
4990
8d stale

Detects lead-lag contagion patterns across asset classes using EHIQ sensor state correlations.

MacroMEDIUM

Dealer Capacity

WITHHELD — falsifier null
--
UNMEASURED
4990
31d stale

The intermediation channel of the rate story: the 30Y swap spread (swap rate minus 30Y treasury yield, bp) as the market-priced measure of dealer balance-sheet capacity. More negative = dealers must be paid more to hold treasuries than to pay fixed = balance-sheet constraint priced. v2 (Eric-stamped 2026-10-03): the v1 falsifier failed as written (2019 repo spike is a repo-market event, not a swap-spread event — declared in the build memo).

Macro

Dining Search Momentum (4-week, Google Trends)

WITHHELD — SHADOW: NH 2026-10-16 re-specified (prereg 38/39); building track record, not counted until admitted
--
UNMEASURED
4991
12d stale

4-week momentum (Google Trends): US search interest in "opentable" and "restaurant reservation", the mean of the last 4 complete weeks against the 4 weeks before them, in percent. Not a seated-diner count and not a year-over-year change. Weekly; holiday and summer search swings are part of the measure. Cuts are frozen in prereg #38 and addendum #39 from the 228 weeks ending 2026-04-11. The display score is 50 + 50 x tanh of half the robust z of the momentum: no floor, no cap.

CryptoMEDIUM

ETH Capitulation Composite Sensor

WITHHELD — mock data published as real: runner builds them with no data provider (eric-stamp ML 2026-10-16)
--
UNMEASURED
4996
28d stale

Composite gauge of 6 capitulation components for ETH market structure.

Crypto

Eth L2 Throughput

--
CALM
4990
10d stale
Crypto

Eth Liquid Staking Ratio

--
CALM
4990
10d stale
Crypto

Eth Validator Exit Queue

WITHHELD — dead input: no BeaconChain API key configured (every reading NULL) (eric-stamp ML 2026-10-16)
--
UNMEASURED
4990
10d stale
FundamentalMEDIUM

Executive Sentiment Divergence

WITHHELD — dead input: Tonality IQ narrative leg never supplied, so divergence is always 0 (eric-stamp QH 2026-10-17)
--
UNMEASURED
4995
8d stale

Divergence between executive transactions and market consensus.

CryptoMEDIUM

GARCH Convergence

WITHHELD — dead input: Python arch module missing (GARCH leg never computes) (eric-stamp ML 2026-10-16)
--
UNMEASURED
4990
8d stale

BTC implied vol mispricing: Deribit DVOL (implied) vs GARCH(1,1)-forecast realized vol from daily prices. Ratio = IV / GARCH forecast; the options-market fear gauge against what realized vol actually justifies.

alternative

GitHub Commit Velocity: BTC

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Engineering velocity proxy via public GitHub commit frequency

alternative

GitHub Commit Velocity: CIFR

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Engineering velocity proxy via public GitHub commit frequency

alternative

GitHub Commit Velocity: ETH

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Engineering velocity proxy via public GitHub commit frequency

alternative

GitHub Commit Velocity: IREN

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Engineering velocity proxy via public GitHub commit frequency

alternative

GitHub Commit Velocity: OPEN

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Engineering velocity proxy via public GitHub commit frequency

alternative

GitHub Commit Velocity: PLTR

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Engineering velocity proxy via public GitHub commit frequency

alternative

GitHub Commit Velocity: PTON

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Engineering velocity proxy via public GitHub commit frequency

alternative

GitHub Commit Velocity: SOL

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Engineering velocity proxy via public GitHub commit frequency

alternative

GitHub Commit Velocity: UPST

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Engineering velocity proxy via public GitHub commit frequency

equity

Innovation Momentum

WITHHELD — dead input: all patent inputs read 0 (eric-stamp ML 2026-10-16)
--
UNMEASURED
4990
17d stale

Innovation momentum indicator from intellectual property activity.

alternative

Job Posting Velocity: BETR

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Hiring velocity proxy via public job board posting counts

alternative

Job Posting Velocity: BITF

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Hiring velocity proxy via public job board posting counts

alternative

Job Posting Velocity: CIFR

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Hiring velocity proxy via public job board posting counts

alternative

Job Posting Velocity: IREN

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Hiring velocity proxy via public job board posting counts

alternative

Job Posting Velocity: NXDR

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Hiring velocity proxy via public job board posting counts

alternative

Job Posting Velocity: OPEN

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Hiring velocity proxy via public job board posting counts

alternative

Job Posting Velocity: PTON

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Hiring velocity proxy via public job board posting counts

alternative

Job Posting Velocity: ROOT

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Hiring velocity proxy via public job board posting counts

alternative

Job Posting Velocity: SANA

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Hiring velocity proxy via public job board posting counts

alternative

Job Posting Velocity: UPST

WITHHELD — SHADOW: LR 2026-10-16 first Neon write of a never-validated alt-data sensor; building track record, not counted until admitted
--
UNMEASURED
4990
11d stale

Hiring velocity proxy via public job board posting counts

equityMEDIUM

SPEC Single-Name

WITHHELD — falsifier fail
--
UNMEASURED
4990
30d stale

The single-name squeeze companion (SPEC-SN): reads retail squeezes at the NAME level — the episode class the broad SPEC composite is structurally blind to (a mean-of-names leg dilutes a one-name surge; pairwise co-movement punishes exactly what a squeeze is). Per name-day on floored flag counts (bjzz_n >= 500; a thin day is ABSENT, never a read): z of QMP-signed imbalance against a trailing-60d baseline excluding the +-10 trading days around the day itself. Instrument value = max-name z + carrier. Raw states QUIET / WARM / SQUEEZING mapped NORMAL / RISING / ELEVATED. Leg 4 = the shelf-readiness governor (Kawa May-2024 framework, Eric-stamped 2026-10-02/03): a SHELF-READY target can dilute itself free, so its squeeze ceiling is capped — FULL caps the state at WARM (z ceiling 2.0); HALF z ceiling 3.0; TOKEN z ceiling 4.0, strength scaling with remaining program capacity from SEC EDGAR primary docs. NO-SHELF / UNKNOWN keep the standard ceiling; blackout is a reported modifier, never a cap input. Companion to spec-composite, NEVER pooled.

Premium Sensors

Unlock all →
Get EHIQ Premium — $249/mo or $1,999/yr

Annual saves 33%. Founding-1000 pricing locked for life.

EventHorizonIQ is an intelligence layer. It provides diagnostic observations, not investment advice, trading signals, or recommendations. All sensor data reflects current conditions at time of recording. Past states do not predict future outcomes. No execution capabilities.