Research Flow
PIT
point-in-time ingestion and availability windows
Back to projects
Independent / Research Infrastructure Project
Quant Research Lab: point-in-time factor research and reproducible backtesting.
Quant Research Lab is a public-facing factor research and backtesting demo covering point-in-time ingestion, signal construction, portfolio formation, evaluation, transaction-cost modeling, benchmark comparison, and attribution.
Key Outcomes
A public demo for empirical research discipline.
Validation
IC
rank information coefficients and factor returns
Risk
Stress
transaction-cost, covariance, exposure, and regime checks
Project Breakdown
Problem, method, system, validation, results, reliability, and research value.
Problem
Backtests can be persuasive while hiding invalid assumptions.
- Lookahead variables, survivorship effects, unstable universes, future price references, and transaction-cost fragility can turn a backtest into a false claim.
- The project needed to make assumptions, availability windows, and validation constraints visible.
Method
The demo follows the full empirical finance research loop.
- Covered point-in-time ingestion, signal construction, portfolio formation, evaluation, transaction-cost modeling, benchmark comparison, and attribution.
- Implemented factors for momentum, value, quality, volatility, liquidity, seasonality, residualization, and cross-sectional normalization.
System / Stack
The stack supports reproducible research artifacts.
- Used Python, FastAPI, pandas, NumPy, SciPy, cvxpy, SQL, Docker, GitHub Actions, Streamlit, experiment manifests, and backtesting workflows.
- Organized each module around explicit assumptions and availability windows.
Validation Methodology
Tearsheets were designed for falsification, not decoration.
- Generated rank ICs, factor returns, turnover, drawdown, exposure decomposition, covariance sensitivity, deflated Sharpe diagnostics, transaction-cost stress tests, and regime breakdowns.
- Compared strategies against benchmarks and attribution surfaces under transaction-cost and regime assumptions.
Results
The project became a teaching artifact for reproducible quant research.
- Built a public-facing factor research and backtesting demo with point-in-time data, signal construction, portfolio formation, evaluation, transaction-cost modeling, benchmark comparison, and attribution.
- Made charts traceable to assumptions, data windows, transformations, validation constraints, code commits, and falsification checks.
Failure Modes / Reliability Checks
The checks target the common ways empirical finance overclaims.
- Tracked lookahead variables, survivorship effects, future price references, unstable universe definitions, duplicate securities, time-window alignment errors, transaction costs, covariance sensitivity, and regime instability.
Why It Matters for Research
It makes model comparison legible under changing assumptions.
- The project mirrors broader research practice: assumptions, data availability, validation constraints, uncertainty, and failure checks should be part of the artifact, not left in memory.