Glossary
Definitions for every term, state, and score used across EventHorizon IQ. New here? Read “Start Here” first — it explains how the whole format fits together.
Start Here — How To Read The Format
EventHorizon IQ is a diagnostic dashboard, not a trading room. It watches markets through sensors, describes the market's current regime, and publishes dated, verifiable calls. Nothing here tells you to buy or sell — it tells you what the data is doing and lets you decide. Four ideas run through everything:
First session: glance at the sensor colors to gauge the mood, check the current regime for context, then read the live calls and the track record. The rest of this page defines every term you'll run into.
Sensor States
Every sensor reports one of these states, and the color is the severity: green is normal, amber is stress building, red and purple are extreme. The four you'll see most often are STABLE (normal), RISING (building), ELEVATED (high), and SEVERE (extreme). One sensor firing is a heads-up; several firing together is the signal that matters.
The normal baseline. Low volatility, balanced risk appetite. The measured asset or indicator is within its typical range. Nothing to act on.
Markets favoring risk assets. Capital flowing into equities, crypto, and high-yield instruments.
No dominant directional bias. Cross-asset flows are roughly even between risk-on and risk-off.
Stress building. The measured indicator is trending toward elevated levels but has not reached critical thresholds.
Capital rotating to safety. Flows moving from equities and crypto into treasuries, gold, and cash.
Crypto markets are leading the macro move. On-chain and exchange data show activity ahead of traditional markets.
Traditional macro indicators are leading. Rate changes, yield moves, or policy shifts driving before crypto responds.
Heightened risk. The measured indicator is significantly above normal and warrants attention.
Extreme stress. Conditions at or near historical extremes. Structural dislocation possible.
Forced or panic selling detected. On-chain or exchange data shows distressed exits at scale.
Extreme bullish sentiment. Leverage, retail inflows, and social metrics at historical highs. Fragile upside.
No directional signal. The sensor does not currently detect conditions that map to a directional state.
How to use: Scan the colors first. Green states need no action — conditions are normal. An amber RISING reading is an early warning to pay attention, not to act. Red ELEVATED and purple SEVERE mean conditions are at or near historical extremes — the moment to review your own exposure and risk, not to chase the move. At true extremes, EHIQ's history shows these often mark contrarian turning points rather than the start of the trend.
Calls, Scoreboard & Track Record
A published, dated prediction: a ticker, a direction (a structural BUY / BULLISH or SELL / BEARISH), and the entry price at the moment it was made. Each call has a clear resolution — it either plays out or it doesn't — and it resolves in the open on the public scoreboard, scored against a fixed forward window (for example, the 30-day forward return).
How to use: A call is a timestamped record, not a tip. Read the entry price and date, then judge the call against where the name trades now. If something is presented as a “call” with no ticker, date, and entry price, it isn't one.
The public, running record of every call and how it resolved — every signal timestamped, every outcome recorded, including the misses. It is built on the immutable ledger, so entries can't be quietly deleted or back-dated. The track record also carries a “What We Got Wrong” section on purpose.
How to use: This is where you audit the whole system. Before trusting any single call, look at the scoreboard to see how the calls have actually done over time — and be suspicious of any track record that only shows winners.
The share of resolved calls that went the predicted direction (for example, an 80% hit rate means 8 of every 10 matured calls resolved correctly). Rates are computed only from calls old enough to have resolved; newer calls stay “pending” until their forward window closes.
How to use: Judge a hit rate over a large sample, not a hot streak, and read it alongside the sample size. A high hit rate on 5 calls means far less than a slightly lower one on 50. The point of the immutable ledger is that the losses can't be hidden.
Scoring Systems
The combined scoring engine. TIQ-ML merges TIPS v2 (qualitative earnings analysis, 60% weight) with a macro-context ML model (40% weight). The ML model uses macro conditions (VIX, yields, sector momentum) to modulate conviction — it never overrides the qualitative direction.
A 0–100 score derived from 20 calibrated language patterns in earnings transcripts. Higher scores indicate stronger structural conviction in the company's trajectory. TIPS is the primary signal; the ML model modulates it.
BULLISH or BEARISH. Derived from qualitative tonality analysis of management language across sequential earnings calls. Not a price prediction — it reflects the structural trajectory of the business.
How to use: Read direction as a multi-quarter thesis, not a short-term timing signal. A BULLISH read can persist while the stock chops sideways for months.
Signal strength as a percentage. Higher conviction means more structural patterns are aligned. A 90% conviction BULLISH signal means nearly all 20 qualitative patterns point in the same direction.
How to use: Conviction is about agreement among the patterns, not certainty of the outcome. Weight the high-conviction reads more heavily and treat the low-conviction ones as tentative — it does not promise the call will work.
The ML model abstains (shows GATED) when its probability is too close to 50/50. This filters out low-confidence predictions. When the gate is active, only the TIPS qualitative score is used.
Inflection Signals
Structural turning points detected from patterns in sequential earnings transcripts. When multiple qualitative patterns converge (constraint dissolution, posture shifts, commercial acceleration), the system flags an inflection signal. These are observations, not trade recommendations.
How to use: An inflection signal says “the story just changed” — it is a prompt to look closer at a name, not an instruction to act. The higher the severity below, the more patterns fired at once.
Market Regimes
A regime is the market's prevailing “weather.” Where a sensor state is one signal's reading, the regime is the overall backdrop those sensors are firing into. It changes slowly — over weeks, not minutes — and the same call means something different depending on which regime you're in.
Calm, low-volatility conditions. Risk is being rewarded, trends are orderly, and stress indicators are quiet. The benign default backdrop.
A healthy risk-on uptrend. Capital is flowing into risk assets and breadth is broadly constructive.
Unstable and two-way. The market is whipsawing and reacting to itself; moves are sharp and mean-reverting. Elevated caution warranted.
Active risk-off. Volatility and credit stress are high and prices are falling. Capital is rotating to safety.
How to use: Read the regime as context, not as a trade. The same bullish call is a tailwind in EXPANSION and a fight-the-tape gamble in STRESS_DRAWDOWN. A regime that is escalating toward stress is your cue to size more conservatively; one that is de-escalating toward calm is the opposite.
General Terms
A persistent market condition that constrains the behavior of assets within it. Regimes are classified by sensors (e.g., VIX regime, Fed policy regime) and change infrequently relative to price.
A classification of a company's multi-quarter trajectory based on qualitative patterns. Arc types range from recovery arcs (Type 1) to cyclical arcs (Type 7) and indicate structural phase, not price direction.
A single signal. An automated classifier that monitors one specific data source (VIX, Treasury yields, on-chain data, etc.) and outputs a state with a confidence level. The sensors are the building blocks everything else is assembled from.
When two markets that usually move together stop agreeing — for example, equities pushing to new highs while credit spreads stay flat, or price rising while on-chain flows fade. The classic read is “the bond market isn’t buying it”: one market is pricing a different reality than the other. How to use: treat a divergence alert as a flag that a move may be unconfirmed. When stocks rally but credit or on-chain data won’t confirm, read the rally as fragile — the slower, more conservative market is often the one that turns out to be right.
The append-only record of all sensor state changes. Every reading is timestamped and cannot be edited or deleted after recording. This allows independent verification of EHIQ's track record.
A suggestion from the AI agent system for a new sensor or coverage expansion. Proposals are reviewed and approved by humans before implementation.
Two sequential earnings transcripts from the same company. The qualitative change between them (what shifted in management language) is the core signal.
EventHorizon IQ 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.