How Market Pulse Intelligence synthesizes raw market data into proprietary composite signals.
Market Pulse Intelligence monitors 100+ data points across six major market domains and combines them into proprietary composite signals. Each signal is designed to highlight specific market conditions that are difficult to see by watching any single data point in isolation. This page provides a high-level overview of our methodology without revealing the specific weightings, thresholds, or formulas that make our signals proprietary.
Design Principles
Every Pulse Signal follows three core principles:
Composite over single-factor. No individual data point is ever enough. Each signal blends multiple indicators to reduce noise and surface durable themes.
Divergence detection. Our signals are designed to flag when different markets are telling conflicting stories — the information edge lives in the divergence, not the consensus.
Real-time + contextual. Every signal combines fast-moving market data (prices, volatility) with slower contextual data (credit conditions, macro indicators, policy expectations).
The Six Pulse Signal Families
Market Pulse Score (Ensemble ML)
A machine learning ensemble combining XGBoost with a momentum/mean-reversion signal. The model is retrained daily on rolling historical data using SHAP feature selection to surface the most predictive features from a pool of 128+ candidates. Outputs include a probability-of-up score, directional label, and the top feature drivers behind each prediction.
Tracked accuracy, cross-validation scores, and 60-day backtest performance are displayed transparently. Past accuracy is not a guarantee of future performance.
Credit Pulse Signals
Designed to capture credit stress before it shows up in headline numbers. Components track:
Fallen Angel Pressure — ratio-based signal from investment-grade spread tiers (BBB relative to AAA) to detect IG-to-HY downgrade risk
Distress Tier Widening — high yield spread dispersion (CCC relative to BB) to flag defaults imminent
War Premium — defense sector outperformance + prediction market probabilities
Contagion Risk — stress-spreading detection across asset classes
Commodity Pulse Signals
Monitors supply/demand stress across commodity complexes:
Energy Stress — oil and natural gas volatility and direction
Precious Metals Flow — gold/silver momentum and gold-vs-equity outperformance
Industrial Demand — Dr. Copper + Baltic Dry shipping as real-economy signals
Ag Pressure — food inflation signal from wheat and corn
Dollar Wind — dollar direction as commodity tailwind or headwind
Supply Disruption Risk — geopolitical overlay on commodity prices
Pulse Predictions (Cross-Platform)
We aggregate prediction market data from Polymarket and Kalshi to produce a unified Prediction Market Risk signal. Our algorithm scans all tracked markets for risk-relevant themes (recession, war, default, tariff, conflict, etc.) and groups them into geopolitical and economic risk categories. This surfaces what the crowd is actually pricing in real time, which often diverges from what traditional markets or news headlines suggest.
News Aggregation & AI Analysis
Six news categories (Breaking, Macro, Credit, Geopolitical, Commodities, Crypto) are aggregated from Google News RSS, CoinDesk, CoinTelegraph, Decrypt, and other free news feeds. Raw headlines are processed through:
Velocity tracking — story volume changes to detect accelerating narratives
High-impact detection — pattern matching for escalation keywords
AI synthesis — Claude Haiku generates concise 2-3 sentence summaries that connect dots across signals
Pulse Briefs
Three times daily (6:30 AM, 12:00 PM, and 5:00 PM ET), we publish "What You Need to Know" executive briefs that synthesize all Pulse Wire summaries, macro calendar events, and market signals into a single digest. Briefs include Top Takeaways, upcoming macro calendar events, and a section-by-section synthesis covering the six major market domains. Briefs are cached once generated to minimize AI compute cost.
Data Update Cadence
Prices (equities, crypto, commodities, currencies) — refreshed every 10 seconds during market hours
Macro and credit data (FRED series) — cached server-side with 5-minute TTL, updates as FRED publishes
Crypto derivatives (OKX) — cached with 60-second TTL
Prediction markets (Polymarket + Kalshi) — cached with 2-minute TTL
News feeds — polled every 10 seconds, deduplicated, sorted newest first
AI summaries — cached 15 to 60 minutes depending on section, to keep compute cost low
Data Sources
All data is sourced from reputable free or paid public APIs:
Yahoo Finance — equities, ETFs, commodities, currencies, global markets
FRED (Federal Reserve Economic Data) — 40+ macro and credit series
ICE BofA indices via FRED — corporate bond spread tiers
Anthropic Claude API — AI-generated summaries and briefs
What Makes These Signals Proprietary
Anyone can look up the VIX, or the 10-year Treasury yield, or BTC funding rates. What makes Market Pulse Intelligence signals proprietary is:
Specific composite weightings — we do not publish the exact weights applied to each component, nor the threshold levels that trigger each label
Cross-domain synthesis — combining data from credit, crypto, macro, and geopolitical sources into unified risk signals that don't exist in any single data provider
Divergence logic — our signals are specifically tuned to flag when markets disagree with each other, which is where most actionable information lives
Continuous refinement — the ML model is retrained daily, and signal thresholds are periodically reviewed against historical outcomes
Important: All Pulse Signals are provided for informational and educational purposes only. They are experimental composite indicators and do not constitute financial advice. Signals may be wrong, may not predict future market movements, and should not be used as the sole basis for any investment decision. See our Disclaimer and Terms of Service for details.