Overview

Workflow

The diagram below shows a general flowchart of the research process, of course since ADRS is built to be composable, users can use the components freely to build their own research workflow too. Hence your use case might slightly differ from this process but for the sake of learning, we will assume this for the rest of the guide.

There are four main components: Data, Alpha, Performance and Report and we will further explore each component in details in later sections.

Data

Handles querying data from upstream data providers and provide tools for pagination, caching, etc. By default, Datasource is used but it is possible to implement your custom handler interacting with other data providers. For your information, Datasource is a data aggregator / provider service at Balaena Quant, note that most of the data are proprietary thus it is recommended that you write your own handler for data that are not freely available.

In addition, it also provides primitives for users to build their own Data Proccessor which is used to handle multiple data sources, for example when an alpha requires hourly data from Glassnode but also requires 15 minute data from Binance then there has to be some sort of upsampling / downsampling, this is where users define their desired behaviour when writing their data processors.

For more details read data.

Alpha

This component provide a base class Alpha for users to extend from such that they can build their own models around the data and to write their desired logic for transforming these model data into a trading signal.

For more details read alpha.

Performance

By default, an alpha only produces the signal at every time step hence the profit and loss (P&L) is not computed at the time of signal generation. Hence, usually the computed signals are then passed into an Evaluator along with the price data to be used where the Evaluator then evaluate the P&L of the alpha at each time step based on the given price.

From the P&L of the alpha, various metrics can be then derived from it for example, the well-known Sharpe Ratio. Of course, it is possible to also derive other metrics such as Calmar Ratio, Max Drawdown, Compound Annual Growth Rate (CAGR), etc.

For more details read performance.

Report

For more details read report.