Blueshift API Documentation

What is Blueshift?

Blueshift is a flexible, event-driven, platform agnostic, trading and backtesting framework for developing systematic investment strategies in Python in a fast and reliable way. This makes developing complex (and simple) strategies easy, and moving a strategy from back-testing to live trading seamless.

Note

Blueshift supports Python 3.6 and newer only. We do not have any planned release for earlier versions of Python.

What it is NOT

Blueshift is NOT meant for any kind of high-frequency trading (HFT). Also, Blueshift is a technical platform and engine for trading and backtesting. It in no way recommends or guarantees the performance of a particular way of trading or investing.

How Blueshift Works

The Blueshift engine, at its core, consists of an event loop designed to run trading algorithms and a set of API functions to support functionalities required for a trading strategy (e.g. fetching tradable instrument details, placing orders or fetching market data etc.).

A Blueshift strategy code does not run directly with the Python interpreter as a regular script would. Instead, the user strategy usually defines a set of event handling functions (or class methods) that determines how the strategy should respond if any event occurs during its run.

The user strategy is read and parsed by the Blueshift engine and is run within its context, in the event loop. The core engine takes care of connecting to the broker and data source(s), running the program and keeping track of the state of the strategy (e.g. trades, orders statuses, profit and loss etc.) automatically. This allows us to focus on the trading logic of the strategy and leave the rest to Blueshift.

Blueshift core engine, together with user defined strategy, forms a Complex Event Processing system (CEP). The Blueshift engine (engine), once launched (in either backtesting or live trading mode), does its own initialization and enters the event processing loop. In this event loop, it continuously looks for incoming events (e.g. new market data arrival or an order fill event etc.), updates its own state and calls the appropriate handler(s) defined by the user strategy at each loop iteration.

The strategy codes (strategy, a Python script or module) written by users define a set of callback functions that are called by the Blueshift engine when the specific event occurs. In addition, users can also use the scheduling functions for finer control.

User defined strategy has no direct access to the event loop or its internal variables. Instead, all interactions with the algo are achieved through a set of API functions (detailed below) and by exposing two special variables -context and data. All data queries are directed through this data object. All algo state related queries (e.g. the cash in the account, current positions etc.) are directed through the context object.

This simple event loop architecture allows blueshift to run any strategies that can be coded, while keeping the interface between the backtesting and live trading consistent.

Blueshift APIs Overview

Blueshift APIs can be divided into below broad categories.

• Callback functions

  The callback functions are the entry points to the algo event loop. These are based on the lifetime events of the strategy. See callback APIs for more on these. All of these callback functions accept the context variable as the first argument and the data variable as the second argument (with the exception of initialize and analyze).

• Scheduling functions

  Scheduling functions allow finer control of event handling by precisely defining repetitive or one-time event handlers. See schedule APIs for more on these. These event handlers also accept two variables - context and data.

• Asset and Data fetching APIs

  Use the current and history method of the data object to fetch current and historical data. Use the symbol function (see assets) to fetch supported assets.

• Trading APIs

  Use the base order function, or its derivatives (like order_target and order_target_percent) for placing orders. Use the cancel_order to cancel an open order. Use get_open_orders to fetch current open orders. For more details, see the trading APIs.

• Pipeline APIs

  Use pipelines for dynamically selecting and filtering the very large asset universe. See pipeline APIs for more details.

• Risk management and other APIs

  A host of risk management APIs to put constraints on the running algo - see risk management APIs for details. Fine tune backtest runs using backtest model selection APIs. Refer to miscellaneous APIs for various useful functions like get_datetime and record.

• Context and Data Objects Attributes

  Use the context object methods for querying the state of the running algo (e.g. querying current portfolio and account states). Use the Data Object methods for fetching current or historical data.

Danger

Backtesting and especially automated trading is far from fire-and-forget. Many things can go wrong (although we try our best to catch and highlight such cases by design). Also there are things beyond control of Blueshift that can go wrong. Like the strategy code going rogue or the datafeed being corrupt or the connectivity unstable. Be careful and remember to take appropriate precautions.