🧠 Smart Money Intelligence

Wall Street consensus, insider activity & hedge fund tracking — all from free data sources

Live Monitor Algo Performance Goldmine Multi-Dimensional
⚠️ Educational tool only. Not financial advice. Data from free public APIs (SEC EDGAR, Finnhub). Signals are informational — always do your own research.
Tracking 12 mega-cap stocks across 5 sectors — why these?
Smart Money Intelligence analyzes a curated universe of 12 high-liquidity, mega-cap stocks chosen for maximum data availability from free sources (SEC 13F filings, Finnhub analyst ratings, Form 4 insider trades). These are among the most widely held and heavily analyzed stocks by institutional investors, making their smart money signals the most meaningful.
AAPLMSFTGOOGL AMZNNVDAMETA JPMBACWMT XOMNFLXJNJ
Tech: AAPL, MSFT, GOOGL, NVDA, META • Consumer: AMZN, NFLX, WMT • Financial: JPM, BAC • Energy: XOM • Healthcare: JNJ
Data sources: SEC EDGAR (13F + Form 4), Finnhub (analyst ratings, insider MSPR), 14 hedge funds tracked (Berkshire, Bridgewater, Citadel, Renaissance, Two Sigma, DE Shaw, Soros, Appaloosa, Baupost, Viking, Millennium, ARK, Third Point, Tiger Global). Updated daily via GitHub Actions.

Top Bullish

--
Awaiting data

Top Bearish

--
Awaiting data

Avg Score

--
12 mega-cap stocks, 5 sectors

Challenger Bot

--
Awaiting data
Rank Ticker Score Direction Confidence Technical Smart Money Insider Analyst Momentum

Loading consensus data...

Loading analyst ratings...
Loading insider activity data...

Recent Insider Trades

Date Ticker Insider Name Title Type Shares Value MSPR

Loading insider trades...

Fund Holdings Matrix

Loading fund holdings...

Conviction Picks (Held by 3+ Funds)


Loading conviction picks...

New Positions


Loading new positions...

Signal Source Performance

Source Total Signals Win Rate Avg 7d Return Avg 30d Return Best Pick Worst Pick

Loading leaderboard...

Fund Performance Rankings

Fund Picks Tracked Avg Return Best Pick

Loading fund rankings...

Head-to-Head: Challenger Bot vs Best Algo

★ Challenger Bot

Win Rate--
PnL--
Sharpe--
Max DD--
Trades--

★ Best Algo

Win Rate--
PnL--
Sharpe--
Max DD--
Trades--

Cumulative PnL Over Time

Challenger Rankings Among All Algos

Rank Algorithm Win Rate PnL Sharpe

Loading rankings...

Recent Challenger Trades

Date Ticker Type Entry Current PnL Status

Loading trades...

Component Accuracy

Which consensus factor is most accurate?

System Analysis DIAGNOSIS

Current Status

• Tracks institutional money flow via SEC filings
• 4 data sources: 13F, Form 4, Analyst Ratings, WSB Sentiment
• 9 database tables for comprehensive tracking
• Powers Challenger Bot (algorithm #20 in Live Monitor)
• Consensus scoring: rule-based weighted formula
• Updated daily via GitHub Actions

Issues Identified

• Challenger Bot 0% WR: 2 trades, 2 losses — FIX DEPLOYED: regime gating + TP/SL ratio
• WSB sentiment noisy: Reddit data is unreliable/manipulated
• 13F data is 45-day delayed: Quarterly filings are stale by the time they appear
• No ML optimization: Consensus weights are arbitrary (30/25/25/20)
• No options flow data: Missing dark pool & unusual activity
• Reddit API rate limits: Restricts real-time sentiment capture

What's Working

• Multi-source institutional data aggregation
• SEC EDGAR filing integration (13F + Form 4)
• Transparent consensus methodology
• Leaderboard ranking across all pillars
• 14 hedge funds tracked (Berkshire, Bridgewater, etc.)
• Automated daily data refresh pipeline

Challenger Bot Performance

Record: 0% Win Rate (2 trades, 2 losses) — FIX DEPLOYED Feb 12
AMZN: -2.49% loss. NVDA: -1.79% loss.

Root Cause: No regime gating — generated BUY signals in bear/neutral markets. Threshold too loose (55), TP/SL ratio 1:1 (no edge).
Fix Applied: (1) Regime gate blocks BUY in bear, SHORT in bull. (2) Threshold raised to 65/35. (3) TP/SL ratio improved to ~2:1 (TP 5%+, SL 2.5-3%). (4) Analyst price target integration for dynamic TP.

Consensus Scoring Methodology

Component Weight Data Source Update Frequency
Analyst Consensus30%Finnhub APIDaily
Insider MSPR25%SEC EDGAR Form 4Daily
13F Institutional25%SEC EDGAR 13FQuarterly
WSB Sentiment20%Reddit APIDaily

System Architecture

Data Collection AUTOMATED
GitHub Actions → 4 Python scripts → MySQL tables.
sec_edgar_13f.py — quarterly institutional holdings
sec_edgar_form4.py — daily insider transactions
wsb_sentiment.py — Reddit mention velocity
smart_money_consensus.py — score aggregation
Database Layer 9 TABLES
sm_analyst_ratings, sm_insider_trades, sm_institutional_holdings, sm_institutional_trades, sm_consensus_scores, sm_fund_performance, sm_fund_rankings, sm_wsb_sentiment, sm_earnings_surprise
API & Frontend 1848 LINES
smart_money.php — unified API with 6 actions (consensus, analyst, insider, institutional, leaderboard, showdown).
smart-money.html — 8-tab dashboard with charts & tables.

Competitor Comparison

Platform Price Key Feature We Have?
Whale Wisdom$50/mo13F analytics + heat mapsYES
Unusual Whales$40/moOptions flow + dark poolsNO
TipRanks$30/moAnalyst accuracy rankingPARTIAL
StocktwitsFreeSocial sentimentYES (WSB)

Database Schema (9 Tables)

sm_analyst_ratings
Wall Street buy/hold/sell + targets
sm_insider_trades
SEC Form 4 buy/sell with MSPR
sm_institutional_holdings
13F quarterly position snapshots
sm_institutional_trades
Derived buy/sell from 13F diffs
sm_consensus_scores
Weighted composite scores (0-100)
sm_fund_performance
Hedge fund return tracking
sm_fund_rankings
Fund leaderboard by performance
sm_wsb_sentiment
Reddit WSB mention velocity
sm_earnings_surprise
Earnings beat/miss history
Research & Future ROADMAP

Academic Foundations

Insider Trading Predictive Power

Lakonishok & Lee (2001): Insiders outperform the market by 4.8% per year. Insider purchases are more informative than sales (sales are often for liquidity, not conviction).
Key insight: Insider buying clusters are strongly predictive. Our MSPR metric captures this, but needs ML weighting to identify the most informative insiders.

13F Herding Behavior

Wermers (1999): Institutional herding — when multiple funds buy the same stock simultaneously — is a strong predictor of future outperformance. Stocks with high institutional momentum outperform by 2-4% over the next quarter.
Our status: We track 14 hedge funds but don't yet quantify herding. Adding herding signals to consensus scoring is a planned improvement.

Analyst Rating Value

Womack (1996): Analyst upgrades produce +5% returns over 6 months; downgrades produce -11% returns. The asymmetry is key — downgrades are more informative than upgrades.
Implication: Our system should weight analyst downgrades more heavily than upgrades in the consensus formula.

Social Sentiment

Chen et al (2014): StockTwits sentiment predicts next-day returns with statistical significance. However, social media data is noisy — signal decays within 1-2 days.
Our status: WSB sentiment is 20% of consensus but is the noisiest component. Reducing its weight or adding sentiment decay is recommended.

Signal Types & Approaches

Insider Buying Clusters
Multiple insiders buying the same stock within 30 days. Historically outperforms by 4.8%/year (Lakonishok & Lee). Strongest signal in our toolkit.
13F Institutional Momentum
Net new positions from top hedge funds. Herding effect amplifies returns when 3+ funds buy simultaneously. 45-day delay is the main limitation.
Analyst Consensus Shifts
Tracking changes in buy/hold/sell distribution over time. Downgrades -11% over 6 months (Womack). Focus on rating changes, not static ratings.
Social Sentiment Velocity
Reddit WSB mention frequency and sentiment polarity. Short-lived signal (1-2 day decay). Best used as a contrarian indicator at extremes.

ML Optimization (Not Yet Implemented)

Priority P1: Learn Optimal Consensus Weights
Current weights are arbitrary: analyst 30%, insider 25%, 13F 25%, WSB 20%. An ML model (gradient boosting or logistic regression) trained on historical consensus scores vs actual returns would learn which signals actually predict future performance. Expected improvement: 20-40% better signal accuracy based on similar studies. Requires minimum 100+ scored predictions with outcomes.

Future Priorities

Priority Improvement Expected Impact Status
P1ML-optimize consensus weights20-40% better signal accuracyPLANNED
P2Add options flow data (dark pool, unusual activity)Capture institutional intentPLANNED
P313D/13G activist filing trackingDetect activist campaigns earlyPLANNED
P4Executive compensation alignment scoringGovernance quality signalPLANNED
P5Sector rotation signals from 13F aggregateMacro sector timingPLANNED
P6Real-time Form 4 filing alertsReduce insider data latencyPLANNED

Benchmark Gap Analysis

Dimension World-Class Ours Gap
Data Sources10+4HIGH
Options FlowYesNoneHIGH
ML OptimizationEnsembleRule-basedHIGH
Update SpeedReal-timeDaily/QuarterlyMED

Sources: Lakonishok & Lee (2001) “Are Insider Trades Informative?”, Wermers (1999) “Mutual Fund Herding”, Womack (1996) “Do Brokerage Analysts' Recommendations Have Investment Value?”, Chen et al (2014) “Wisdom of Crowds: StockTwits Sentiment”, SEC EDGAR documentation, Finnhub API docs.

System Analysis RESEARCH

How Smart Money Intelligence works:
Aggregates institutional and insider data from multiple sources into a single consensus score (0–100). Three data pillars: Analyst Ratings (Wall Street consensus from Finnhub — Buy/Hold/Sell distribution, average price target), Insider Trading (SEC Form 4 filings via EDGAR — insider buy/sell ratio, MSPR score), and 13F Institutional Holdings (SEC EDGAR quarterly filings — which hedge funds are buying/selling). The consensus score weights all three pillars to produce an actionable signal.

Data pipeline:
• Python scripts run via GitHub Actions (weekdays 6AM + Sunday 9AM EST)
• SEC EDGAR 13F parser: Extracts position changes from latest quarterly filings
• Form 4 parser: Real-time insider buy/sell tracking with MSPR (months since purchase ratio)
• Finnhub API: Analyst recommendations and price targets
• WSB sentiment: Reddit WallStreetBets mention velocity (experimental)

Current status:
• This is a research-only system — it provides intelligence, not direct trade signals
• The Challenger Bot (algorithm #20 in Live Monitor) uses consensus scores for automated trading (≥70 = BUY, ≤30 = SHORT)
• Consensus accuracy tracking is active but insufficient data for statistical significance yet
• L vs O (#1 PICK) is the only system with verified positive returns

Data Sources & Transparency LIVE DATA

Where the numbers come from:
• Consensus scores: smart_money.php?action=consensus — combined analyst + insider + institutional score
• Analyst data: smart_money.php?action=analyst — Finnhub recommendation trends
• Insider trades: smart_money.php?action=insider — SEC Form 4 filings via EDGAR
• 13F holdings: smart_money.php?action=institutional — quarterly hedge fund filings
• Leaderboard: smart_money.php?action=leaderboard — top-scored stocks across all pillars
• GitHub Actions: smart-money-tracker.yml — automated data refresh schedule
Smart Money Intelligence — Part of Find Toronto Events — Data refreshes every 5 minutes