Quant Research Engine

A quantitative research platform for backtest overfitting detection, prediction market analysis, options analytics, and information diffusion modeling.

Status

System D — Backtest Overfitting Detector — shipped. Statistics (Sharpe, PSR, DSR), the backtest engine, the strategy generator, the experiment runner, overfitting analysis, and the overfitting research study are all on master.

System A — Prediction Market Mispricing Engine — in progress. Market data contract, Polymarket and Kalshi read-only adapters, dataset persistence and replay, the order book model with a mechanical execution simulator, and same-market/cross-venue arbitrage detection with cost and market-identity models are shipped — work runs through Phase 2.6 (arbitrage normalization). See docs/market-data.md, docs/order-book.md, docs/arbitrage.md, docs/cross-venue-arbitrage.md.

Systems B (Hawkes Information Diffusion) and C (Breeden–Litzenberger Risk-Neutral Distribution) — not started.

This is research/paper-trading software only. No live trading. No real money. No financial advice.

Planned Systems

System	Name	Status
A	Prediction Market Mispricing Engine	In Progress — data, order book, and arbitrage stacks shipped (through Phase 2.6)
B	Hawkes Information Diffusion Model	Planned
C	Breeden–Litzenberger Risk-Neutral Distribution	Planned
D	Backtest Overfitting Detector	Shipped

Development

Setup

# Clone and enter the project
cd quant-research-engine

# Create virtual environment
python -m venv .venv
.venv\Scripts\activate  # Windows
# source .venv/bin/activate  # macOS/Linux

# Install in development mode
pip install -e ".[dev]"

# Copy environment config
cp .env.example .env

Linting and Type Checking

# Format code
ruff format src/ tests/

# Lint
ruff check src/ tests/

# Type check
mypy src/

Testing

The suite runs under pytest (testpaths = ["tests"] in pyproject.toml): 34 test files covering unit tests plus Hypothesis property-based tests in tests/property/.

# Run all tests
pytest

# Run with coverage
pytest --cov=quant_engine

# Run specific test file
pytest tests/unit/test_random.py -v

CI (.github/workflows/ci.yml, on every push to master and every pull request) is the gate and the source of truth:

Check	Command
Lint	ruff check .
Type check	mypy — strict mode (strict = true in pyproject.toml, pinned mypy==1.20.2 + scipy-stubs==1.18.1.1)
Tests	pytest

Subsystem design and behavior are documented under docs/ — see Documentation.

Architecture

See docs/architecture.md for the full architecture documentation.

Key principle: Core/domain code must NOT depend on concrete database, API, filesystem, or frontend implementations.

Documentation

Topic	Document
Architecture and current phase	docs/architecture.md
Statistics (Sharpe, PSR, DSR)	docs/statistics.md
Backtest engine	docs/backtesting.md
Strategy generation	docs/strategy_generation.md
Experiments	docs/experiments.md
Overfitting analysis	docs/overfitting.md
Overfitting research study	docs/research/overfitting-study.md
Market data contract	docs/market-data.md
Polymarket adapter	docs/providers/polymarket.md
Kalshi adapter	docs/providers/kalshi.md
Order book and execution	docs/order-book.md
Same-market arbitrage	docs/arbitrage.md
Arbitrage costs	docs/arbitrage-costs.md
Cross-venue arbitrage	docs/cross-venue-arbitrage.md
Market identity	docs/market-identity.md

License

MIT — see LICENSE.