Factor Development
Factor Storage
Stores factors data from across research teams into a centralized library — handling high-frequency and low frequency factors across equities, futures, options and more. Flexible partitioning by time × security or time × factor name supports both high-throughput batch writes and fast cross-sectional and time-series queries across the full market and long historical periods.
Benchmarked at hundreds of GB ingested in minutes and full-market single-factor updates in seconds.
Featured Tutorials & Best Practices:
Mid-to-High Frequency Factor Computation
A Turing-complete scripting language, vectorized functions, and streaming engines support cross-sectional, time-series and cross-asset factor logic — unifying microstructure, technical, style, and fundamental factors in a single framework.
Factor logic can be encapsulated as UDFs and shared across strategy groups. Built-in libraries including WorldQuant 101 Alpha, Technical Analysis Indicator Library, and Cross-Sectional Asset Pricing Factor provide a head start on research.
The unified stream-batch architecture ensures the same factor logic runs on historical data for backtesting and on live streams for live trading — no duplicate codebases, no logic drift.
Featured Tutorials & Best Practices:
- Best Practices for Factor Computation
- Stream-Batch Unified Factor Development:
- Quick Guide to a Unified Batch-Stream Computing Platform
- Calculating High-Frequency Factors in Real-Time
- Real-time scenarios include: Real-time Rank of Price Rate of Change](https://docs.dolphindb.com/en/Tutorials/realtime_ROC_rankings.html)
- Cases for Factor Implementation: General Financial Factors, ETF Options Implied Volatility and Greeks, Daily Factors of Mutual Fund
- Factor Calculation Solutions Compared:
- Python + HDF5 vs. DolphinDB
- Python + File Storage vs. DolphinDB
Factor Evaluation
DolphinDB computes IC, Rank IC, stratified portfolio returns, turnover, and stability metrics across industry, style, and market regimes using vectorized window functions and distributed computing. Hundreds of candidate factors can be screened and evaluated in hours rather than days. Single-factor return attribution and risk profiling help research teams identify alpha sources and failure scenarios, informing downstream multi-factor modeling.
Featured Tutorials & Best Practices:
- High-Performance Factor Evaluation with DolphinDB
- Application of Alphalens in DolphinDB: Practical Factor Analysis Modeling
Multi-Factor Modeling
DolphinDB aligns selected factors with market, financial, position, trading, and risk data by time and security in a single step — generating feature tables and label sets ready for modeling.
- Multi-Factor Framework: Linear weighting, cross-sectional regression, and IC-weighted models run directly in-database, producing scores or expected returns for stock selection and timing.
- Risk Factors & Portfolio Optimization: A built-in risk model framework supports factor exposure analysis, return attribution, and portfolio optimization under position, industry/style neutrality, and exposure constraints.
- ML & Deep Learning: XGBoost, AI Dataloader, LibTorch, and a built-in ML framework enable large-scale model training and real-time inference — all within DolphinDB.
Featured Tutorials & Best Practices:
- XGBoost Plugin Application: Feature Engineering for Stock Volatility Prediction
- AI Dataloader and LibTorch Tools: DolphinDB Application in Deep Learning: Real-Time Stock Volatility Prediction
Automated Factor Discovery
DolphinDB's Shark GPLearn is a genetic algorithm-based framework that uses symbolic regression to automatically generate Alpha expressions from massive historical data — purpose-built for institutions with large-scale factor libraries.
Compared to open-source gplearn, Shark GPLearn adds GPU acceleration, integrates 2,000+ built-in DolphinDB operators, and natively handles three-dimensional panel data (time × stock × factor) — dramatically compressing the generate-screen-backtest iteration cycle.
Request a free trial → https://dolphindb.com/product#shark
Featured Tutorials & Best Practices:
Factor Lab
A fully integrated platform for the entire factor lifecycle — from data management and research to evaluation and strategy backtesting. Every factor's computation logic, data dependencies, historical performance, and applied strategies are automatically tracked, with full lineage tracing and impact analysis.
Teams share a unified view of IC, IC_IR, long-short returns, and other key metrics across the full factor library — enabling performance comparison, portfolio optimization, and collaborative research without duplicated effort.
Request a free trial → https://dolphindb.com/product#starfish
Featured Tutorials & Best Practices: