Market + multi-source news

Daily OHLCV bars for ten macro instruments, plus roughly 130 headlines per cycle aggregated from three independent news feeds — all normalized into one shape and de-duplicated across sources.

• Macro listeners: oil, gold, silver, dollar, 10Y yield, VIX, SPX, Nasdaq, Dow, nat gas
• Yahoo Finance per-ticker headlines (SPY, QQQ, AAPL, NVDA…)
• Google News RSS — one query per event theme (CPI, Fed, OPEC, jobs…)
• Alpaca News API (Benzinga) when API keys are configured
• On-disk price caching prevents rate-limiting

Price thresholds + headline lexicon

Raw data becomes normalized events two ways: a macro instrument moving past its per-asset daily threshold, or a headline matching the keyword lexicon with word-boundary precision.

• Per-instrument thresholds (oil ±2%, VIX ±8%, SPX ±1%…)
• 15 news event types: CPI, Fed hawkish/dovish, jobs, OPEC, tariffs, FDA, AI capex…
• Direction-aware, with magnitude captured as % daily change
• Word-boundary regex — no false matches on partial words

Events → confidence-scored signals

Active events are evaluated against a declarative 38-rule rulebook spanning eight market themes. Every rule is transparent and auditable — every signal traces back to the exact events that fired it.

• Single-trigger and compound rules (e.g. Jobs miss + Fed dovish → TLT)
• Compound rules require all triggers at once and earn a conviction bonus
• Confidence 0–99 = base score + magnitude bonus + corroboration bonus
• Duplicate firings consolidated; signals below 55 auto-suppressed

Price structure for every ticker

Every actionable signal gets one year of daily candles analyzed for structure before any trade is proposed.

• Swing highs/lows identified with a 5-bar fractal window
• Nearby levels clustered (1.2% tolerance) → nearest support & resistance
• ATR%, volume vs 20-day average, last-candle pattern (engulfing, hammer, doji…)

A complete, risk-managed ticket

Signals are never a naked BUY/SELL — each ships as a concrete proposal anchored to the price structure found in Stage 4. The inverse logic applies to SELL setups.

• Entry at current price; stop beyond nearest S/R with a half-ATR buffer
• Target 1 and Target 2 at the next structural levels
• Risk-to-reward computed; below 1.5× flagged LOW_RR
• Quality rating on every ticket: OK / LOW_RR / NO_LEVELS

611 symbols on demand

Any S&P 500 stock or top US ETF can be analyzed from the search bar — even when no rule targets it directly.

• Sector-sensitivity matrix scores the symbol against every active event
• Net score → BUY / SELL / NEUTRAL bias with confidence
• Tailwind vs headwind event breakdown for transparency
• Direct rule hits count as strong corroborating evidence
• Same S/R engine produces a full trade setup

Continuous background loop

The server re-scans all market and news feeds every 90 seconds (configurable) and re-runs the full pipeline over the active event set — API reads stay near-instant because a live snapshot is always ready.

Announced once, never re-fired

Every event has a stable identity — one per instrument per day for market moves, a headline fingerprint for news — so nothing is announced twice. Stale events expire automatically (market 8 h, news 6 h), so the board never goes stale.

Pushed to the browser

Fresh snapshots stream over Server-Sent Events: the LIVE indicator shows connection state, newly arrived events and signals flash on arrival, and the stream reconnects automatically if the connection drops.

Near-real-time, not tick-level

The engine refreshes every 90 seconds and price events come from daily-bar changes, not intraday ticks. Signals are swing- and position-trade oriented — not built for day-trading.

Transparent heuristics, no black box

All 38 rules are human-authored and auditable — every signal traces back to its exact events and rule. Confidence scores are formula-based, not yet backtested; there is no ML model making opaque calls.

Research only — no orders placed

APEX is fully read-only: it proposes, you decide. There is no brokerage connection or position tracking. Backtesting, alerting, and paper-trade integration are the next phases on the roadmap.