--- name: alphaear-strategy description: "Score trading setups using AlphaEar multi-factor analysis (momentum, volume, sentiment). Use when evaluating entry/exit signals." domain: trading --- ## Skill Definition **Name:** `alphaear-strategy` **Pattern:** `trading/alphaear-strategy` **Description:** Multi-signal trading strategy combining news aggregation, sentiment analysis, options flow, and predictive modeling. Generates entry/exit signals with confidence scoring. --- ## Implementation Four-phase signal synthesis engine combining news aggregation, sentiment layering, predictive modeling, and options flow analysis. ### Phase 1: Real-Time News Aggregation **Multi-Source Intelligence Gathering** ```python news_sources = { "financial_media": ["Bloomberg", "Reuters", "WSJ", "CNBC"], "social_sentiment": ["Twitter/X", "Reddit", "StockTwits"], "prediction_markets": ["Polymarket", "Kalshi"], "regional_sources": ["Cailian (China)", "Economic Times (India)"] } ``` **Transmission Chain Analysis** - Map news propagation speed across sources - Identify "first movers" vs "lagging indicators" - Calculate information diffusion velocity - Detect semantic divergence (same event, different interpretations) ### Phase 2: Sentiment Layering **FinBERT Sentiment Scoring** ```python sentiment_pipeline = { "model": "FinBERT", "scoring_range": (-1.0, +1.0), "confidence_threshold": 0.75, "time_decay": "exponential_24h" } ``` **Multi-Horizon Sentiment** - Immediate reaction (0-1 hour) - Short-term sentiment (1-24 hours) - Medium-term bias (1-7 days) - Structural shift detection (7+ days) **Sentiment Divergence Signals** - Bullish news + bearish price action = caution - Bearish news + accumulation = opportunity - Neutral news + high volume = catalyst pending ### Phase 3: Kronos Predictive Model **Time-Series Forecasting with News Adjustments** ```python kronos_model = { "base": "LSTM_sequence_model", "features": ["price", "volume", "volatility", "sentiment"], "prediction_horizon": ["1h", "4h", "1d", "1w"], "confidence_intervals": True, "news_aware": True } ``` **Signal Evolution Tracking** - Strengthen: Prediction aligns with unfolding reality - Weaken: Model confidence declining - Falsify: Thesis invalidated by new data ### Phase 4: Options Flow Integration **Unusual Activity Detection** ```python options_signals = { "volume_threshold": 2.5 * avg_daily_volume, "sweep_detection": True, "block_trade_filter": volume > 1000, "otm_percentage": 0.10, # 10% OTM "expiration_window": "30d" } ``` **Options-Price Divergence** - High call buying + flat stock = bullish setup - High put buying + rising stock = bear trap - IV expansion without price move = event pending --- ## Signal Generation Framework Composite signal scoring system that weights news, social, options, technical, and prediction layers into a single actionable score. ### Composite Signal Score ```python def calculate_alpha_signal(ticker): # Component scores (0-100) news_score = aggregate_news_sentiment(ticker) social_score = analyze_social_sentiment(ticker) options_score = detect_unusual_flow(ticker) technical_score = evaluate_setup_quality(ticker) prediction_score = kronos_forecast_confidence(ticker) # Weighted composite weights = { 'news': 0.25, 'social': 0.20, 'options': 0.25, 'technical': 0.15, 'prediction': 0.15 } composite = sum(score * weights[key] for key, score in scores.items()) return { 'signal': composite, 'direction': 'LONG' if composite > 65 else 'SHORT' if composite < 35 else 'NEUTRAL', 'confidence': min(composite, 100-composite) if composite != 50 else 0 } ``` ### Signal Thresholds ``` 80-100: STRONG BUY - High conviction, size appropriately 65-79: MODERATE BUY - Good setup, standard sizing 50-64: WEAK BUY - Edge present but marginal 35-49: WEAK SELL - Mild negative edge 20-34: MODERATE SELL - Clear bearish signals 0-19: STRONG SELL - High conviction short ``` --- ## Usage Examples Practical workflows for full analysis, signal monitoring, and event-driven setups. ### Example 1: Full Analysis Workflow ```python # Run complete AlphaEar analysis analysis = alphaear_analyze("NVDA") # Output includes: # - News aggregation with sentiment # - Social media trend analysis # - Options flow anomalies # - Kronos price prediction # - Composite signal score if analysis.signal_score > 75: position_size = portfolio_value * 0.05 # 5% max entry = current_price stop = entry * 0.95 # 5% stop target = entry * 1.15 # 15% target ``` ### Example 2: Signal Monitoring ```python # Monitor multiple positions portfolio = ["AAPL", "TSLA", "NVDA", "AMD"] signals = {} for ticker in portfolio: signals[ticker] = alphaear_analyze(ticker) # Alert on signal degradation for ticker, signal in signals.items(): if signal.evolution == "WEAKEN": alert(f"{ticker}: Signal weakening, review position") elif signal.evolution == "FALSIFY": alert(f"{ticker}: Thesis invalidated, consider exit") ``` ### Example 3: Event-Driven Setup ```python # Pre-earnings analysis ticker = "AMZN" catalyst_date = get_next_earnings_date(ticker) days_to_catalyst = (catalyst_date - today).days if days_to_catalyst <= 7: setup = alphaear_analyze( ticker, focus="catalyst_setup", include_options=True ) if setup.options_signal == "unusual_call_activity": # Market positioning bullish direction = "LONG" structure = "call_spread" ``` --- Load `references/trading-checklist.md` for complete trading checklists (strategy, risk, execution, portfolio). ## Integration Points **Cross-Skill Dependencies** - `trading/investing-algorithm-framework` - Prediction market sentiment overlay - `trading/black-edge` - Alternative data integration - `research/trendradar` - Trend detection correlation - `marketing/social-media-engagement` - Social sentiment analysis **Data Source Requirements** - News APIs (NewsAPI, Bloomberg API) - Social sentiment (Twitter API, Reddit) - Options data (Unusual Whales, Cboe) - Price data (yfinance, Polygon) --- ## Risk Management Signal quality filters, position sizing algorithms, and risk-adjusted allocation rules. ### Signal Quality Filters ```python # Avoid false signals filters = { "minimum_volume": 1000000, # $1M daily volume "minimum_market_cap": 1e9, # $1B minimum "max_spread_percent": 0.5, # 0.5% max spread "earnings_buffer": 2, # Days before/after earnings "news_recency": 24 # Hours } ``` ### Position Sizing ```python def size_position(signal_score, portfolio_value): base_size = portfolio_value * 0.02 # 2% base # Scale by signal strength if signal_score >= 80: multiplier = 2.5 # 5% position elif signal_score >= 65: multiplier = 1.5 # 3% position else: multiplier = 1.0 # 2% position return base_size * multiplier ``` --- ## Output Format ```yaml alphaear_report: ticker: "NVDA" analysis_timestamp: "2025-05-04T14:30:00Z" news_analysis: headline_count: 47 sentiment_score: 0.72 key_themes: - "AI demand acceleration" - "Data center spending" - "Supply chain constraints" transmission_velocity: "fast" social_sentiment: reddit_score: 0.65 twitter_score: 0.78 stocktwits_score: 0.81 trending_hashtags: ["#NVDA", "#AI", "#semiconductors"] options_flow: unusual_activity: True call_put_ratio: 2.3 top_volume_strikes: [900, 950, 1000] sweep_detected: True kronos_prediction: 1h_forecast: "+1.2%" 4h_forecast: "+2.1%" 1d_forecast: "+3.5%" confidence: 0.74 composite_signal: score: 82 direction: "LONG" confidence: "HIGH" evolution: "STRENGTHEN" recommendation: action: "BUY" entry: "Current market" stop_loss: "-5%" take_profit: "+15%" position_size: "5% portfolio" time_frame: "1-2 weeks" risk_factors: - "Overall market volatility elevated" - "Semiconductor sector rotation risk" - "Earnings in 3 weeks" ``` --- ## Triggers - `/alphaear analyze ` - Full multi-signal analysis - `/alphaear scan ` - Find best opportunities in sector - `/alphaear monitor ` - Real-time signal tracking - `/alphaear backtest ` - Historical signal performance --- **Note:** This skill requires access to real-time data feeds and sentiment APIs. Free tiers may have limitations. For production use, consider premium data subscriptions. **Donation:** Support development → https://www.tip.md/oyi77 ## Evidence Standards (Non-Negotiable) | Tier | Type | Weight | Verification | AlphaEar Examples | |------|------|--------|--------------|-------------------| | **T1** | Primary source | 1.0 | Direct URL + timestamp + hash | SEC filings, earnings call transcripts, exchange order book data, Cboe options flow raw data, central bank policy statements | | **T2** | Factual secondary | 0.7 | Cross-reference 2+ independent sources | Bloomberg/Reuters news feeds, Unusual Whales aggregated flow, FinBERT sentiment scores, Polygon price/volume data, Kronos model output (validated against historical backtest) | | **T3** | Opinion/social | 0.3 | Flag as "speculative" in all outputs | Twitter/X trending sentiment, Reddit/StockTwits posts, analyst price targets, newsletter tips, YouTube commentary, prediction market odds (Polymarket/Kalshi) | **Rules:** 1. No actionable recommendation (buy/sell/short/hedge) on T3-only evidence 2. Composite signal score >65 for longs or <35 for shorts requires at least 50% T1/T2 weighted evidence across the 4 signal layers 3. Every T3 claim must be paired with a T1/T2 disconfirming evidence search before inclusion 4. Always disclose evidence composition (T1/T2/T3 percentage) in the output report 5. Kronos predictions alone are T2 evidence at most; they require T1 confirmation from price/volume data ## Anti-Bias Checklist (Run Before Every Recommendation) Six cognitive traps to evaluate before every recommendation to ensure evidence-driven decisions. ### 6 Cognitive Traps - [ ] **Confirmation bias** -- Did I actively seek news and data that contradicts the current signal direction? If bullish, did I search for bearish catalysts? - [ ] **Anchoring** -- Am I over-weighting the first headline or sentiment reading I saw? Have I re-evaluated after scanning all 4 signal layers independently? - [ ] **Recency bias** -- Am I projecting the last 24-48 hours of sentiment forward without checking 30-day and 90-day historical context for this ticker? - [ ] **Herd mentality** -- Is social sentiment simply reflecting the crowd? Have I checked whether options flow and Kronos diverge from the consensus direction? - [ ] **Sunk cost** -- Am I defending a prior position or thesis by overweighting signals that confirm it? Would I take the same trade if I had no existing position? - [ ] **Overconfidence** -- Is my conviction score calibrated to evidence quality? A composite of 75 built on 60% T3 evidence is weaker than a composite of 65 built on 70% T1/T2 evidence ## When NOT to Use - When you have no access to real-time news APIs (NewsAPI, Bloomberg) or options data feeds (Unusual Whales, Cboe) -- the signal synthesis engine cannot function without live data - When analyzing assets with less than $10M average daily volume -- illiquid instruments produce unreliable sentiment and options flow signals - During earnings blackout periods for the target ticker -- options flow, sentiment, and news signals distort around earnings dates, producing false positives - When only a single signal layer is available (e.g., sentiment only, no options or Kronos data) -- multi-signal convergence is the core thesis; single-source signals lack sufficient edge - When the user needs pure technical analysis without news or sentiment overlay -- use `fin-equity-technical` instead - When analyzing pre-IPO or recently listed stocks (<30 days) -- insufficient historical data for Kronos predictions and options flow patterns - When regulatory filings or compliance tasks are the primary goal -- use `fin-compliance-kyc` instead - When the task is personal budgeting, tax preparation, or non-market financial planning -- outside this skill's domain - When a more appropriate specialized skill exists (e.g., `trading/black-edge` for dark pool intelligence, `trading/investing-algorithm-framework` for prediction markets) ## Common Rationalizations | Rationalization | Reality | |---|---| | "News is already priced in" | Research shows 65%+ of post-news drift occurs in the first 4 hours; retail and algo digestion lag creates tradeable windows | | "Sentiment is too noisy to be useful" | Multi-source sentiment aggregation (FinBERT + social + options) filters ~70% of noise; divergence between sentiment and price action is itself a signal | | "Options flow can be hedging, not directional" | Unusual sweep activity filtering (volume >2.5x avg, block trades >1000 contracts, OTM focus) isolates speculative positioning from delta-neutral hedging | | "Kronos model confidence is high, skip the other signals" | No single layer is sufficient -- even at 0.9 confidence, Kronos can miss regime changes that news/sentiment catches; convergence across layers is what creates alpha | | "Just give me the signal, skip the verification" | Skipping verification leads to acting on single-source signals with 30-40% win rates vs. multi-signal convergence at 55-65% win rates | | "I'll size up since the composite score is 85+" | Position sizing must still respect portfolio risk limits; a single trade should never exceed 5% of portfolio regardless of signal strength | | "The stock is up 10% today, the signal must be right" | Chasing momentum without checking options flow and Kronos evolution confirms hindsight bias, not forward edge | | "I don't need the full 4-phase analysis for this" | Truncating phases removes the diversification benefit of multi-signal synthesis; each skipped layer increases false positive probability by 15-25% | ## Red Flags - Signal originates from only one of the four layers (news, sentiment, options, Kronos) without convergence from at least one other layer - Options flow directly contradicts sentiment direction (e.g., heavy put buying alongside bullish sentiment) -- this divergence requires resolution before acting - Kronos model confidence falls below 0.4 -- below this threshold, predictions have no statistical edge over random - News sentiment score exceeds 0.9 or drops below -0.9 -- extreme readings often mean the move has already happened - Social sentiment diverges from financial media sentiment by more than 0.4 -- suggests either retail-FII conflict or coordinated manipulation - Composite signal score between 45-55 (the "no edge" zone) -- trades in this range have negative expected value after transaction costs - Options sweep detected but IV is declining (unusual activity + falling IV = likely hedging, not speculation) - Headline count spikes above 3 standard deviations from 30-day mean -- information overload degrades signal quality - Signal evolution tracking shows WEAKEN or FALSIFY but position is still held -- thesis invalidated, action required - User requests to skip a signal phase or "just use sentiment" -- every phase skipped increases false positive rate by 15-25% - Position size recommendation exceeds 5% of portfolio value regardless of signal strength ## Verification After completing this skill, confirm: - [ ] All 4 signal layers analyzed (news aggregation, sentiment, options flow, Kronos prediction) - [ ] Composite score >65 for longs or <35 for shorts -- no trades in the 45-55 "no edge" zone - [ ] Stop-loss and take-profit levels set before entry - [ ] Position size respects the 5% portfolio max regardless of signal strength - [ ] Signal direction confirmed by at least 2 of 4 layers (convergence check) - [ ] No red flags from the list above are active for the target ticker - [ ] Earnings buffer respected (no new positions within 2 days of earnings) - [ ] Kronos confidence >0.4 and evolution status is STRENGTHEN or neutral (not WEAKEN or FALSIFY) - [ ] Evidence composition disclosed (what percentage of the signal is T1 vs T2 vs T3) - [ ] Anti-Bias Checklist completed with all 6 traps addressed **Cross-reference:** For complete multi-asset analysis, also see `financial/all-in-one-finance` and `financial/wolf-finance`. For dark pool intelligence, see `trading/black-edge`. For prediction market sentiment, see `trading/investing-algorithm-framework`.