Common data types in algorithmic trading
Algo traders rely on several data categories to generate signals and execute strategies. The choice depends on strategy type, from technical rule-based systems to AI-driven models that leverage alternative datasets.
Primary data sources:
- Price and volume: Time series of open, high, low, close (OHLC), and traded volume are fundamental.
- Order book and level-2 data: Depth of market information useful for execution strategies and market-making.
- Fundamental data: Financial statements, earnings, and macroeconomic indicators for longer-term or value-oriented models.
- News and sentiment: Headlines, press releases, and social media sentiment feeds for event-driven strategies.
- Alternative data: Web scraping, satellite imagery, credit card spending, and other nontraditional indicators used by some quant funds.
Quality considerations:
- Frequency: Tick-level data offers precision for high-frequency strategies, while daily bars may suffice for swing strategies.
- Cleanliness: Missing data, bad ticks, and corporate actions (splits, dividends) must be handled properly.
- Latency: Real-time strategies require low-latency feeds; delayed data can be inadequate for execution.
Practical list of tasks for traders:
- Choose data that matches your horizon and instruments.
- Clean and normalize data, adjusting for splits and dividends.
- Store historical data for backtesting and live caches for execution.
- Monitor vendor reliability and plan for redundancy.
Using the right data at the right quality level is crucial: poor inputs will undermine even well-designed algorithms.