Event Prediction

AI That Tells Company Stories, Then Predicts What Happens Next

Back to Foresight

We Read Company Stories Like a Detective Novel

Our AI reads every SEC filing for a company up to a specific date, then predicts what will happen in the next 6-12 months. And it explains why.

1

Company in Trouble

The Story (January 2024):
  • 📉 Month 1: Violates loan agreement
  • 🔍 Month 2: Accountants find internal control problems
  • 📊 Month 3: Credit rating downgraded
AI Prediction:
75% chance of bankruptcy
within 6 months
What Actually Happened: Bankruptcy filed August 2024
AI's Reasoning:
"Covenant violation followed by internal control issues and credit downgrade is a classic distress pattern. Historical data shows 75% of companies with this sequence default within 6 months."
2

Successful Turnaround

The Story (September 2023):
  • ⚠️ Month 1: Accounting problems disclosed
  • ✅ Month 2: Issues fixed, regulatory approval granted
  • 💰 Month 3: Refinances debt at better terms
  • 👔 Month 4: New leadership team appointed
AI Prediction:
70% chance of turnaround
within 12 months
What Actually Happened: Returned to profitability 2024
AI's Reasoning:
"Quick remediation of accounting issues followed by debt refinancing and leadership change indicates strong turnaround momentum. Companies showing this pattern succeed 70% of the time."
🔑 The Big Difference
The AI Explains Its Reasoning
Not just predictions - you get the "why" behind every forecast

What Can We Predict?

Three types of predictions that matter to investors

Mergers & Acquisitions

Predict which companies will be acquired or merge within 6 months

Use Case:
Identify acquisition targets before the market notices. Position ahead of the announcement premium.

Bankruptcy & Distress

Predict which companies will default, delist, or file bankruptcy within 6 months

Use Case:
Early warning system for portfolio risk. Exit positions or short companies before collapse.

Successful Turnarounds

Predict which struggling companies will successfully recover within 12 months

Use Case:
Find distressed value opportunities. Buy companies the market has written off but will recover.

Why We Use Text-Based AI (Not Number Crunchers)

Traditional models struggle with new event types. Our LLM adapts automatically.

❌ Traditional Numerical Models

  • ✗
    Can't handle new events
    Model breaks when a new type of event appears in the system
  • ✗
    Misses the narrative
    Only sees numbers, not the story of what's happening
  • ✗
    Can't explain decisions
    Black box - you never know why it made a prediction
  • ✗
    Requires retraining
    Every time you add new event types, rebuild from scratch

✅ Our Text-Based LLM

  • ✓
    Adapts to new events automatically
    Understands language - new event types work immediately
  • ✓
    Understands the company story
    Reads events like a narrative - sees temporal patterns and context
  • ✓
    Explains every prediction
    Tells you WHY it made each forecast - builds trust with investors
  • ✓
    Future-proof system
    Add new event types without retraining - just works

The Time-Based Story Advantage

📖
Reads Events in Sequence
Understands "problem → fix → recovery" vs "problem → more problems → collapse"
⏰
Timing Matters
3 problems in 3 months is different than 3 problems in 3 years
🎯
Context is King
Same event means different things for different companies in different situations

How Good Is It?

We beat typical hedge fund performance by a significant margin

93-96%
Accurate When It Predicts
When the model makes a call, it's right 19 out of 20 times
30x
Better Than Random
Predictive power is 30 times stronger than guessing
6-12mo
Advance Warning
Predict events 6-12 months before they happen

How We Rank vs Industry

Our Model (Turnarounds) Strong Positive
Correlation: 0.31 (top quartile performance)
Our Model (Distress) Moderate Positive
Correlation: 0.26 (above average)
Typical Hedge Fund Weak Positive
Correlation: 0.05-0.15 (industry benchmark)
What This Means: Our predictions are 2-6x more correlated with actual outcomes than typical hedge fund models. In finance, small edges compound to massive advantages.

Making It Even Better

V11 upgrade: 10x more training data for 20-40% better predictions

V8 CURRENT ✅ Live & Validated

Current System

  • • 45,000 training examples (3 years of data)
  • • 2022-2024 timeframe
  • • Limited economic diversity (bull market only)
  • • Fewer rare events (50-300 per type)
  • • Performance: Good, but room to improve
V11 PLANNED 🎯 Dec 2025

Upgraded System

  • • 300K-800K training examples (10x more data!)
  • • 2005-2024 timeframe (20 years)
  • • Full economic cycles (2008 crisis, COVID, bull & bear markets)
  • • 10x more rare events (better pattern learning)
  • • Performance: 20-40% improvement expected

Why More Historical Data Makes Us Better

10x
More Rare Events
See enough bankruptcies and turnarounds to learn real patterns
3
Market Cycles
2008 crisis, COVID crash, bull markets - model learns to handle any environment
Complete
Story Arcs
See full distress → recovery cycles, not just snapshots