07. Constructing Financial State Spaces Part 1

Understanding State Space and Feature Vectors in Financial Markets

• State Space:

  • A simplified representation of the underlying environment.
  • Made of feature vectors.

• Feature Vector Basics:

  • Numerical representation of data at a given timestep.
  • Ordered collection of numbers.

• Vector Construction:

  • Derived from raw data such as prices and volumes.
  • Ordered collection of raw data, calculated features, and/or technical indicators.

• Raw Data:

  • Price & volume data: open, high, low, close, trade volume.
  • Feature vectors should always have at least one raw price data point, for the bot to buy and sell based on.

• Feature Calculation:

  • Compute complex features and technical indicators from raw data.

  • Popular Features and Indicators:

    • Returns: Percentage change in price.
    • Moving Averages: Trend direction.
    • Momentum Indicators: Momentum of price trends.

• Feature Selection Impact:

  • Determines learned trading policy - since we are learning a model-free policy, the types of data you allow the bot to see will heavily determine the type of policy it learns.

Understanding the integration of these elements aids in building effective financial analysis models. Consider both feature selection and mathematical calculations for optimal results.