Trading Signals

Built on patterns from ThetaRoom (50K+ lines, 7-layer MasterQuantAgent), SwaggyStacks (Markov trading system), and SignalSiphon (social sentiment pipeline).

1. Identify regime → Markov 7-state model (see reference/markov-regime.md)
2. Route by asset → Options? Stocks? Crypto? Commodities? VIX? Forex?
3. Apply methodologies → Confluence scoring with regime-weighted fusion
4. Size the position → Risk management (max 2% per trade, 15% portfolio)
5. Explain the why → Educational mode: every signal comes with reasoning

Confluence score:

• 0.7-1.0: High conviction → execute with full position
• 0.4-0.7: Moderate → wait for more confluence or reduce size
• 0.0-0.4: No trade → stay patient

Quick options analysis:

Ticker + Strike + Expiry → Greeks profile → Strategy fit → Risk/reward → Go/No-go

Asset Class Router

When Multiple Assets Interact

Many real trades span asset classes. Examples:

• "VIX spiked, should I adjust my SPY iron condor?" → vix-volatility.md + options-strategies.md + risk-management.md
• "Gold is rallying, what does that mean for crypto?" → commodities.md + correlation analysis + markov-regime.md
• "My AAPL calls are deep ITM before earnings" → options-trading.md + equities.md + risk-management.md

Signal → Why → Context → Action

Example: "The iron condor makes sense here because IV rank is at 78% — that's top quartile, meaning options premiums are historically expensive. You're selling that rich premium. Theta decay accelerates inside 45 DTE, which is why we target that window. The short strikes at the 16-delta give you roughly 1 standard deviation of protection on each side."

Principles:

• Explain the market structure driving the signal (regime, vol environment, flow)
• Connect Greeks to real P&L impact ("your theta is -$45/day, meaning you earn $45 if nothing moves")
• Reference historical patterns when relevant ("Bitcoin post-halving typically enters Bull Quiet regime within 6 months")
• Flag what could go wrong and why ("if VIX breaks 35, the regime shifts to crisis mode and your iron condor wings are at risk")

The foundation — 5 TA methodologies with regime-weighted confluence scoring.

Confluence Detection

class ConfluenceAnalyzer:
    """Regime-weighted methodology fusion — from ThetaRoom MasterQuantAgent"""

    REGIME_WEIGHTS = {
        'trending_up':   {'elliott': 0.30, 'turtle': 0.30, 'fib': 0.20, 'wyckoff': 0.15},
        'trending_down': {'elliott': 0.30, 'turtle': 0.30, 'fib': 0.20, 'wyckoff': 0.15},
        'ranging':       {'fib': 0.35, 'wyckoff': 0.30, 'elliott': 0.20, 'turtle': 0.05},
        'volatile':      {'fib': 0.30, 'wyckoff': 0.30, 'elliott': 0.20, 'turtle': 0.10},
    }

Score → Action:

• 0.7-1.0: High conviction entry (full position)
• 0.4-0.7: Wait for more confluence (half position or watch)
• 0.0-0.4: No trade (patience pays)

MasterQuantAgent Ensemble (ThetaRoom v1)

7-layer weighted voting for highest-conviction decisions:

8-Node Trading Pipeline (ThetaRoom v2)

Scanner → Volatility → Greeks → Risk → Entry → Position → Execution → Exit

Each node maps to a LangGraph agent. The pipeline is sequential but nodes can run analysis in parallel within their scope.

Cost-Effective Model Routing

Key code references:

• Options Greeks: theta-room/backend/nautilus/greeks_actor.py
• 12 options strategies: swaggy-stacks/backend/app/strategies/options/
• Markov 7-state: swaggy-stacks/backend/app/methodologies/bitcoin/
• Risk config: thetaroom/thetaroom/config.py (ThetaRoomConfig)
• Sentiment: signal-siphon/backend/analyzer/sentiment_analyzer.py
• Brokers: theta-room/backend/brokers/ (Alpaca, Binance, IBKR, Deribit, Polymarket)

Data sources in your stack:

• Polygon.io / Massive — real-time stocks, options, crypto, forex
• yfinance — historical OHLCV
• Alpaca SDK — stock/options/crypto trading + data
• Binance — spot, futures (USD-M, COIN-M), options
• Deribit — BTC options (~90% market share)
• CoinGecko — crypto prices (free tier)
• NautilusTrader — execution engine (ThetaRoom v2)

Technical Analysis Methodologies:

• reference/elliott-wave.md — Wave rules, halving supercycle, crypto adaptation, targets
• reference/turtle-trading.md — Donchian channels, ATR sizing, pyramiding
• reference/fibonacci.md — Levels, golden pocket, Greeks-Fib fusion, on-chain enhanced
• reference/wyckoff.md — Phase state machines, VSA, composite operator
• reference/markov-regime.md — 7-state model, transition probabilities, regime-based signals

Options & Volatility:

• reference/options-trading.md — Greeks (1st + 2nd order), Black-Scholes, IV surface, vol smile, GEX, pin risk
• reference/options-strategies.md — 25+ strategies: income, directional, volatility, advanced multi-leg
• reference/vix-volatility.md — IV rank, VIX term structure, crisis thresholds, regime integration

Asset Classes:

• reference/equities.md — Sector rotation, scanner patterns, Kelly allocation, earnings plays
• reference/commodities.md — Gold/Silver (COT, seasonal, dollar correlation), Oil (inventory, OPEC, contango)
• reference/forex.md — Carry trade, rate differentials, PPP, central bank policy, COT positioning

Cross-Cutting:

• reference/sentiment-signals.md — Social filtering, multi-model consensus, noise detection
• reference/risk-management.md — Position sizing, portfolio Greeks gates, drawdown limits, smart execution
• reference/pattern-recognition.md — Candlestick + chart patterns
• reference/swarm-consensus.md — Multi-LLM voting system
• reference/chinese-llm-stack.md — Cost-optimized Chinese LLMs for trading

Workflow & Backtesting:

• reference/daily-trading-workflow.md — /loop + Desktop scheduled tasks for pre-market, open, intraday, EOD
• reference/backtesting-patterns.md — Walk-forward, Monte Carlo, ensemble, combinatorial alpha discovery

Emit Outcome Sidecar

As the final step, write to ~/.claude/skill-analytics/last-outcome-trading-signals.json:

{"ts":"[UTC ISO8601]","skill":"trading-signals","version":"2.1.0","variant":"default",
 "status":"[success|partial|error]","runtime_ms":[estimated ms from start],
 "metrics":{"signals_generated":[n],"tickers_analyzed":[n],"frameworks_applied":[n]},
 "error":null,"session_id":"[YYYY-MM-DD]"}

Use status "partial" if some stages failed but results were produced. Use "error" only if no output was generated.