🚀 Core Pillar I: Danelfin-Style AI Win Rate Scoring, Earnings AI Analysis & Executive Insider Radar

#

Release Version: v7.2.0
Core Components: core/ai_winrate_engine.py, core/post_earnings_analyzer.py, core/insider_radar.py
Tags: AI Alpha · Multi-Factor Win Rate · Earnings Audit · Insider Risk Radar

📖 Introduction: From Rule-Based Models to AI Alpha Probability Profiling

#

Traditional trading systems often rely on single technical indicators or lagging financial statements, leaving them vulnerable to market noise and false breakouts. The release of V7.2.0 marks the transition of Trade OS into the era of AI Stock Selection & Multi-Factor Alpha Probability Profiling.

By combining a Danelfin-style 1-10 AI Win Rate Engine, Fiscal.ai grade post-earnings call AI analysis, and an SEC Form 4 Executive Insider Radar, the system quantitatively evaluates the probability of an asset outperforming SPY/QQQ over the next 3 months alongside comprehensive risk protection.

🎯 I. 1-10 AI Win Rate Scoring Engine

#

The win rate engine AIWinRateEngine (core/ai_winrate_engine.py) synthesizes multi-source heterogeneous data into a continuous score from 1.0 to 10.0, mapping directly to statistical outperformance probabilities.

1.1 Four-Factor Weighted Scoring Formula

#

The overall score is calculated from four independent sub-factor weighted scores:

$$\text{Overall Score} = 0.35 \times S_{\text{tech}} + 0.30 \times S_{\text{fund}} + 0.20 \times S_{\text{micro}} + 0.15 \times S_{\text{insider}}$$

# Core scoring logic (core/ai_winrate_engine.py)
overall_score = round(
    (valid_tech * 0.35) + 
    (valid_fund * 0.30) + 
    (valid_micro * 0.20) + 
    (insider_score * 0.15), 
    1
)
overall_score = float(np.clip(overall_score, 1.0, 10.0))

• Technical Score $S_{\text{tech}}$ (35%): Evaluates 20/50/200-day SMA bullish alignments, price above 200SMA, and 20-day momentum.
• Fundamental & Consensus Score $S_{\text{fund}}$ (30%): Evaluates YoY revenue growth (>20% bonus), net profit margins, and Wall Street analyst recommendation ratings.
• Micro Volume/Price Score $S_{\text{micro}}$ (20%): Measures recent 5-day volume relative to 20-day average, penalizing high-volume breakdowns (ret_5 < -2% and ratio > 1.2 deducts 2.5 pts).
• Executive Insider Score $S_{\text{insider}}$ (15%): Parses SEC EDGAR filings, awarding 9.0 for insider accumulation and capping at 2.0 for aggressive dumps.

1.2 Outperformance Probability Mapping & Ratings

#

The system maps overall_score directly to 3-month outperformance probabilities:

🎙️ II. Fiscal.ai Grade Post-Earnings AI Analysis & KPI Audit

#

Earnings season provides rich Alpha opportunities but carries concentrated tail risk.

2.1 Post-Earnings Transcript AI Analysis

#

PostEarningsAnalyzer (core/post_earnings_analyzer.py) automatically fetches management call transcripts post-earnings, analyzing Q&A tone confidence (Confidence vs. Hesitation) and auditing industry-specific KPIs:

• SaaS / Software: Extracts ARR (Annual Recurring Revenue) growth and NRR (Net Retention Rate).
• AI Hardware / Semiconductors: Audits Data Center revenue share, CapEx guidance, and Gross Margin trends.

2.2 Three-Stage Event Risk Protection

#

Working with check_earnings_alerts.py, the system establishes a T-7d / T-1d / T-0d progressive risk cutoff. Before market close on T-1d, position sizes and implied volatility are checked to prevent earnings black swan events from severely damaging portfolio capital.

🕵️‍♂️ III. SEC Form 4 Executive Insider Radar

#

InsiderRadar (core/insider_radar.py) resolves information asymmetry for retail investors.

3.1 EDGAR Data Stream Parsing

#

The system continuously monitors SEC EDGAR Form 4 filings, identifying transactions by CEOs, CFOs, and 10%+ key shareholders (Code P Buy vs. Code S Sell).

# Insider transaction status rule
if net_insider_volume > 0 and buy_count >= 2:
    status = "INSIDER-ACCUMULATION"  # Executive buying signal
elif sell_count >= 3 and sell_ratio > 0.8:
    status = "INSIDER-DUMP"          # Concentrated executive dump

When INSIDER-DUMP is detected, the AI Win Rate Engine drops the insider sub-score to 2.0, pulling down overall win rate and triggering console security alerts.

💡 Summary & Architectural Significance

#

Through AIWinRateEngine (core/ai_winrate_engine.py), Trade OS translates raw financial data into intuitive win probabilities. Whether in live market scanning or the Unified Action Terminal, traders immediately view clear win rate badges and multi-factor breakdowns, elevating decision speed and win rate consistency.