Market Signals

🔗 Live Application: getstreetinsights.com

Stock sentiment tracking with credibility-weighted source analysis. Automatically detects ticker mention spikes, captures predictions with reasoning, validates outcomes, and builds source credibility scores over time.

Architecture

Flow:

1. Scanner detects frequency spikes for tickers across platforms (Twitter, Reddit, news)
2. Capture extracts sentiment, price targets, timeframes, and reasoning
3. Attribution identifies and profiles the source
4. Evaluation scores reasoning quality using equity analyst frameworks (Lynch, Munger)
5. Validation compares predictions to actual outcomes over time
6. Credibility updates source scores based on track record
7. Signals provides credibility-weighted sentiment aggregation

Database Schema

Core Tables

• tickers - Stock symbols being tracked
• sources - Analysts, influencers, publications making predictions
• mentions - Raw detected mentions of tickers
• predictions - Structured predictions with reasoning
• validations - Outcomes validating prediction accuracy
• mention_frequency - Daily aggregated mention counts for spike detection

Key Features

• Automatic credibility scoring (70% accuracy + 20% volume + 10% reasoning quality)
• Source track record tracking
• Reasoning quality evaluation based on equity analyst frameworks
• Spike detection for ticker mentions

Tech Stack

• Frontend: Vite + React 18 + TypeScript
• UI: Tailwind CSS + shadcn/ui
• Database: Supabase (PostgreSQL)
• AI Analysis: XAI Grok (source reasoning evaluation)
• Data Sources: Twitter/X API, Reddit API, Alpha Vantage

Setup

1. Clone and install:

   npm install

2. Configure environment:

   cp .env.example .env
   # Fill in your API keys

3. Set up Supabase:

   • Create a new Supabase project
   • Run the schema: supabase-schema.sql
   • Copy URL and anon key to .env

4. Run development server:

   npm run dev

Evaluation Prompts

The system uses three equity analyst frameworks to evaluate source reasoning quality:

1. The Lynch Pitch (Bull Case)

Evaluates if the source:

• States a clear investment thesis
• Cites real data and documents
• Explains the business model and competitive advantages
• Identifies specific catalysts
• Considers market position and risks

2. The Munger Invert (Bear Case)

Evaluates if the source:

• Considers structural weaknesses
• Identifies balance sheet risks
• Examines competitive threats
• Addresses management credibility
• Acknowledges what could go wrong

3. Management Analysis

Evaluates if the source:

• Reviews actual management guidance vs reality
• Examines financial statement trends
• Identifies strategy execution and capital allocation
• Considers insider behavior

Key Concepts

Credibility Score (0-100):

• 70% prediction accuracy
• 20% prediction volume (capped at 50)
• 10% reasoning quality

Reasoning Quality (0-1):

• Data discipline (do they cite sources?)
• Transparency (do they admit uncertainty?)
• Framework usage (do they use sound reasoning?)

Source Types:

• Individual (retail traders, individuals)
• Publication (Seeking Alpha, Benzinga, MarketWatch)
• Analyst Firm (Goldman Sachs, Morgan Stanley)
• Influencer (high-follower accounts)

Repurposing trend-weaver

The scanner component is adapted from trend-weaver, pivoting from "content generation" to "mention detection + sentiment aggregation":

trend-weaver → market-signals:

• Tweet tracking → Mention tracking
• Engagement metrics → Credibility metrics
• Topics → Tickers
• Content generation → Source evaluation

Project Status

• Database schema designed
• Scanner built (adapt from trend-weaver)
• Sentiment capture pipeline
• Source evaluation with equity analyst prompts
• Historical validation system
• Dashboard UI

License

Private - Boxford Partners