Documentation Hub

Comprehensive documentation for the BRISMA implied risk premia framework. Academic papers, tutorials, and technical references.

Academic Research

Research Paper (69 pages)

Implied Risk Premia for Factors: Theory, Estimation, and Applications. Comprehensive academic treatment with real Fama-French data.

PDF Paper

Fama-French Data Pipeline

Complete documentation of the data pipeline using Kenneth French Data Library (July 1963 - December 2023, 726 months).

Pipeline Docs

Getting Started

Getting Started Guide

Quick start instructions for running the BRISMA pipeline. Covers environment setup, data loading, and basic usage.

Quickstart

CLI Setup Guide

New joiner onboarding guide. How to set up the development environment and run the pipeline from command line.

Setup

Theory & Background

Introduction to Implied Risk Premia

Comprehensive 17-section introduction covering the theoretical foundations of implied returns extraction from portfolio weights.

Theory

Implied Returns Methodology

Detailed methodology for extracting implied expected returns using mean-variance inverse optimization.

Methodology

Factor Premia Documentation

Factor model decomposition and extraction of implied factor risk premia from asset-level implied returns.

Methodology

Reference

Mathematical Wiki

Complete mathematical documentation with formulas, derivations, and proofs. Sections 14-15 cover Bantleon methods in detail.

Reference

Architecture Diagrams

System architecture documentation with data flow diagrams, module relationships, and component interactions.

Technical

Tutorials

Tutorial: Julian's Methods

Step-by-step tutorial covering Bantleon's M1, M2, and Hybrid calibration methods with exercises and worked examples.

Tutorial

Key Concepts Reference

Implied Returns

mu* = lambda * Q * w + rf

Extract expected returns from observed portfolio weights, covariance matrix, and risk aversion.

Lambda M1 (Market)

lambda = ln((1+y_10Y)/(1+rf)) / beta_ref

Calibrate lambda using 10-year bond yield spread. Used when R-squared > 0.6.

Lambda M2 (Historical)

lambda = sum(w_t * f_t)

Exponentially weighted historical factor returns. Used when R-squared < 0.3.

Hybrid Blend

mu = R^2 * mu_M1 + (1-R^2) * mu_M2

Blend methods based on R-squared regime indicator for smooth transitions.

Factor Model

r = B * f + epsilon

Decompose returns into systematic (factor) and idiosyncratic components.

Factor Premia

pi = (B'B)^-1 * B' * (mu* - rf)

Extract implied factor risk premia from asset-level implied returns.

External Resources

🔗

GitHub: Digital-AI-Finance

📝

Example 1 Walkthrough