Skip to content

Valuation Models

The framework employs multiple valuation approaches to provide diverse perspectives on stock value. The dashboard displays 7 active models, each with different strengths, assumptions, and data requirements.

Active Dashboard Models

These are the models currently shown on the dashboard, listed in display order:

ML Ensemble

  • AutoResearch - 5-model ensemble predicting peak 2-year returns

Peak Return Prediction (Opportunistic)

Fixed-Horizon Return Prediction

  • GBM 1y - Gradient boosted machine, 1-year return prediction
  • GBM 3y - Gradient boosted machine, 3-year return prediction

Classic Valuation

Model Comparison

Model Type Predicts Coverage Data Required Best For
AutoResearch ML ensemble (5 models) Peak 2y return High Fundamental snapshots + price Ranking stocks by upside potential
GBM Opp 1y Gradient boosting Peak 1y return ~52% 8+ quarters Tactical trades, timing
GBM Opp 3y Gradient boosting Peak 3y return ~52% 8+ quarters Multi-year opportunities
GBM 1y Gradient boosting 1-year return ~52% 8+ quarters Portfolio construction
GBM 3y Gradient boosting 3-year return ~52% 8+ quarters Long-term ranking
DCF Fundamental Absolute fair value ~98% 1+ quarters Stable cash flow companies
RIM Fundamental Absolute fair value ~85% Book value data Banks, asset-heavy companies

How to Use Multiple Models

The dashboard displays all available models for each stock. Consider:

  1. Convergence: When multiple models agree on upside, confidence is higher
  2. Divergence: Indicates model assumptions may not fit the business
  3. ML vs Fundamental: GBM/AutoResearch models rank stocks relatively; DCF/RIM provide absolute fair values
  4. Time Horizons: Compare 1y vs 3y predictions to assess near-term vs long-term opportunity

Understanding Predictions

  • Fair Value: Estimated intrinsic value per share (DCF, RIM) or implied price from predicted return (ML models)
  • Upside %: Predicted return or margin of safety vs current price
  • Confidence: Model-specific reliability indicator (rank percentile for ML models)

Retired Models

The following models were previously available but have been removed from the dashboard:

  • GBM Lite 1y/3y: Simplified GBM with 59 features and 2-quarter minimum. Removed for being overoptimistic. See archived docs.
  • Enhanced DCF: DCF variant accounting for dividend policy. See archived docs.
  • Multi-Stage DCF: Multiple growth phase DCF. Removed (85% failure rate). See archived docs.
  • Growth DCF: Separated growth vs maintenance CapEx. Removed (85% failure rate). See archived docs.
  • Simple Ratios: Market multiples valuation. Removed (broken). See archived docs.

For detailed information on each active model, click the links above.