Skip to content

A high-performance algorithmic trading platform and event-driven backtester

License

Notifications You must be signed in to change notification settings

RiverRo/nautilus_trader

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

codacy-quality codecov pythons pypi-version pypi-format Downloads discord

Branch Version Status
master version build
develop version build
Platform Rust Python
Linux (x86_64) 1.63.0+ 3.8+
macOS (x86_64) 1.63.0+ 3.8+
Windows (x86_64) 1.63.0+ 3.8+

Introduction

NautilusTrader is an open-source, high-performance, production-grade algorithmic trading platform, providing quantitative traders with the ability to backtest portfolios of automated trading strategies on historical data with an event-driven engine, and also deploy those same strategies live, with no code changes.

The platform is 'AI-first', designed to develop and deploy algorithmic trading strategies within a highly performant and robust Python native environment. This helps to address the parity challenge of keeping the Python research/backtest environment, consistent with the production live trading environment.

NautilusTraders design, architecture and implementation philosophy holds software correctness and safety at the highest level, with the aim of supporting Python native, mission-critical, trading system backtesting and live deployment workloads.

The platform is also universal and asset class agnostic - with any REST, WebSocket or FIX API able to be integrated via modular adapters. Thus, it can handle high-frequency trading operations for any asset classes including FX, Equities, Futures, Options, CFDs, Crypto and Betting - across multiple venues simultaneously.

Features

  • Fast - C-level speed through Rust and Cython. Asynchronous networking with uvloop
  • Reliable - Type safety through Rust and Cython. Redis backed performant state persistence
  • Portable - OS independent, runs on Linux, macOS, Windows. Deploy using Docker
  • Flexible - Modular adapters mean any REST, WebSocket, or FIX API can be integrated
  • Advanced - Time in force IOC, FOK, GTD, AT_THE_OPEN, AT_THE_CLOSE, advanced order types and conditional triggers. Execution instructions post-only, reduce-only, and icebergs. Contingency order lists including OCO, OTO
  • Customizable - Add user defined custom components, or assemble entire systems from scratch leveraging the cache and message bus
  • Backtesting - Run with multiple venues, instruments and strategies simultaneously using historical quote tick, trade tick, bar, order book and custom data with nanosecond resolution
  • Live - Use identical strategy implementations between backtesting and live deployments
  • Multi-venue - Multiple venue capabilities facilitate market making and statistical arbitrage strategies
  • AI Agent Training - Backtest engine fast enough to be used to train AI trading agents (RL/ES)

Alt text

nautilus - from ancient Greek 'sailor' and naus 'ship'.

The nautilus shell consists of modular chambers with a growth factor which approximates a logarithmic spiral. The idea is that this can be translated to the aesthetics of design and architecture.

Why NautilusTrader?

  • Highly performant event-driven Python - native binary core components
  • Parity between backtesting and live trading - identical strategy code
  • Reduced operational risk - risk management functionality, logical correctness and type safety
  • Highly extendable - message bus, custom components and actors, custom data, custom adapters

Traditionally, trading strategy research and backtesting might be conducted in Python (or other suitable language) using vectorized methods, with the strategy then needing to be reimplemented in a more event-drive way using C++, C#, Java or other statically typed language(s). The reasoning here is that vectorized backtesting code cannot express the granular time and event dependent complexity of real-time trading, where compiled languages have proven to be more suitable due to their inherently higher performance, and type safety.

One of the key advantages of NautilusTrader here, is that this reimplementation step is now circumvented - as the critical core components of the platform have all been written entirely in Rust or Cython. This means we're using the right tools for the job, where systems programming languages compile performant binaries, with CPython C extension modules then able to offer a Python native environment, suitable for professional quantitative traders and trading firms.

Why Python?

Python was originally created decades ago as a simple scripting language with a clean straight forward syntax. It has since evolved into a fully fledged general purpose object-oriented programming language. Based on the TIOBE index, Python is currently the most popular programming language in the world. Not only that, Python has become the de facto lingua franca of data science, machine learning, and artificial intelligence.

The language out of the box is not without its drawbacks however, especially in the context of implementing large performance-critical systems. Cython has addressed a lot of these issues, offering all the advantages of a statically typed language, embedded into Pythons rich ecosystem of software libraries and developer/user communities.

What is Cython?

Cython is a compiled programming language which aims to be a superset of the Python programming language, designed to give C-like performance with code that is written in Python - with optional C-inspired syntax.

The project heavily utilizes Cython to provide static type safety and increased performance for Python through C extension modules. The vast majority of the production code is actually written in Cython, however the libraries can be accessed from both pure Python and Cython.

What is Rust?

Rust is a multi-paradigm programming language designed for performance and safety, especially safe concurrency. Rust is blazingly fast and memory-efficient (comparable to C and C++) with no runtime or garbage collector. It can power mission-critical systems, run on embedded devices, and easily integrates with other languages.

Rust’s rich type system and ownership model guarantees memory-safety and thread-safety deterministically — eliminating many classes of bugs at compile-time.

The project increasingly utilizes Rust for core performance-critical components. Python language binding is handled through Cython, with static libraries linked at compile-time before the wheel binaries are packaged, so a user does not need to have Rust installed to run NautilusTrader. In the future as more Rust code is introduced, PyO3 will be leveraged for easier Python bindings.

Architecture (data flow)

Architecture

Quality Attributes

  • Reliability
  • Performance
  • Modularity
  • Testability
  • Maintainability
  • Deployability

Integrations

NautilusTrader is designed in a modular way to work with 'adapters' which provide connectivity to data publishers and/or trading venues - converting their raw API into a unified interface. The following integrations are currently supported:

Name ID Type Status Docs
Betfair BETFAIR Sports Betting Exchange status Guide
Binance BINANCE Crypto Exchange (CEX) status Guide
Binance US BINANCE Crypto Exchange (CEX) status Guide
Binance Futures BINANCE Crypto Exchange (CEX) status Guide
FTX FTX Crypto Exchange (CEX) status Guide
FTX US FTX Crypto Exchange (CEX) status Guide
Interactive Brokers IB Brokerage (multi-venue) status Guide

Refer to the Integrations documentation for further details.

Installation

From PyPI

We recommend running the platform with the latest stable version of Python, and in a virtual environment to isolate the dependencies.

To install the latest binary wheel from PyPI:

pip install -U nautilus_trader

To install numpy and scipy on ARM architectures such as MacBook Pro M1 / Apple Silicon, this stackoverflow thread is useful.

From Source

Installation from source requires the Python.h header file, which is included in development releases such as python-dev. You'll also need the latest stable rustc and cargo to compile the Rust libraries.

It's possible to install from source using pip if you first install the build dependencies as specified in the pyproject.toml. However, we highly recommend installing using poetry as below.

  1. Install rustup (the Rust toolchain installer):

  2. Enable cargo in the current shell:

    • Linux and macOS:
      source $HOME/.cargo/env
      
    • Windows:
      • Start a new PowerShell
  3. Install poetry (or follow the installation guide on their site):

    curl -sSL https://install.python-poetry.org | python3 -
    
  4. Clone the source with git, and install from the projects root directory:

    git clone https://github.com/nautechsystems/nautilus_trader
    cd nautilus_trader
    poetry install --no-dev
    

Refer to the Installation Guide for other options and further details.

Versioning and releases

NautilusTrader is currently following a bi-weekly beta release schedule. The API is becoming more stable, however breaking changes are still possible between releases. Documentation of these changes in the release notes are made on a best-effort basis.

Branches

  • master branch will always reflect the source code for the latest released version
  • develop branch is normally very active with frequent commits and may contain experimental features. We aim to maintain a stable passing build on this branch

The current roadmap has a goal of achieving a stable API for a 2.x version. From this point we will follow a formal process for releases, with deprecation periods for any API changes.

Makefile

A Makefile is provided to automate most installation and build tasks. It provides the following targets:

  • make install -- Installs the package using poetry
  • make build -- Runs the Cython build script
  • make clean -- Cleans all none source artifacts from the repository
  • make docs -- Builds the documentation HTML using Sphinx
  • make pre-commit -- Runs the pre-commit checks over all files

Examples

Indicators and strategies can be developed in both Python and Cython (although if performance and latency sensitivity are import we recommend Cython). The below are some examples of this:

  • indicator example written in Python
  • indicator examples written in Cython
  • strategy examples written in both Python and Cython
  • backtest examples using a BacktestEngine directly

Docker

Docker containers are built using a base python:3.10-slim with the following image variant tags:

  • nautilus_trader:latest has the latest release version installed
  • nautilus_trader:develop has the head of the develop branch installed
  • jupyterlab:develop has the head of the develop branch installed along with jupyterlab and an example backtest notebook with accompanying data

The container images can be pulled as follows:

docker pull ghcr.io/nautechsystems/<image_variant_tag>

You can launch the backtest example container by running:

docker pull ghcr.io/nautechsystems/jupyterlab:develop
docker run -p 8888:8888 ghcr.io/nautechsystems/jupyterlab:develop

Then navigate to the backtest_example.ipynb and run it!

Minimal Strategy

The following is a minimal EMA Cross strategy example which just uses bar data. While trading strategies can become very advanced with this platform, it's still possible to put together simple strategies. First inherit from the Strategy base class, then only the methods which are required by the strategy need to be implemented.

class EMACross(Strategy):
    """
    A simple moving average cross example strategy.

    When the fast EMA crosses the slow EMA then enter a position at the market
    in that direction.

    Cancels all orders and closes all positions on stop.
    """

    def __init__(self, config: EMACrossConfig):
        super().__init__(config)

        # Configuration
        self.instrument_id = InstrumentId.from_str(config.instrument_id)
        self.bar_type = BarType.from_str(config.bar_type)
        self.trade_size = Decimal(config.trade_size)

        # Create the indicators for the strategy
        self.fast_ema = ExponentialMovingAverage(config.fast_ema_period)
        self.slow_ema = ExponentialMovingAverage(config.slow_ema_period)

        self.instrument: Optional[Instrument] = None  # Initialized in on_start

    def on_start(self):
        """Actions to be performed on strategy start."""
        # Get instrument
        self.instrument = self.cache.instrument(self.instrument_id)

        # Register the indicators for updating
        self.register_indicator_for_bars(self.bar_type, self.fast_ema)
        self.register_indicator_for_bars(self.bar_type, self.slow_ema)

        # Get historical data
        self.request_bars(self.bar_type)

        # Subscribe to live data
        self.subscribe_bars(self.bar_type)

    def on_bar(self, bar: Bar):
        """Actions to be performed when the strategy receives a bar."""
        # BUY LOGIC
        if self.fast_ema.value >= self.slow_ema.value:
            if self.portfolio.is_flat(self.instrument_id):
                self.buy()
            elif self.portfolio.is_net_short(self.instrument_id):
                self.close_all_positions(self.instrument_id)
                self.buy()
        # SELL LOGIC
        elif self.fast_ema.value < self.slow_ema.value:
            if self.portfolio.is_flat(self.instrument_id):
                self.sell()
            elif self.portfolio.is_net_long(self.instrument_id):
                self.close_all_positions(self.instrument_id)
                self.sell()

    def buy(self):
        """Users simple buy method (example)."""
        order: MarketOrder = self.order_factory.market(
            instrument_id=self.instrument_id,
            order_side=OrderSide.BUY,
            quantity=self.instrument.make_qty(self.trade_size),
        )

        self.submit_order(order)

    def sell(self):
        """Users simple sell method (example)."""
        order: MarketOrder = self.order_factory.market(
            instrument_id=self.instrument_id,
            order_side=OrderSide.SELL,
            quantity=self.instrument.make_qty(self.trade_size),
        )

        self.submit_order(order)

    def on_stop(self):
        """Actions to be performed when the strategy is stopped."""
        # Cleanup orders and positions
        self.cancel_all_orders(self.instrument_id)
        self.close_all_positions(self.instrument_id)

        # Unsubscribe from data
        self.unsubscribe_bars(self.bar_type)

    def on_reset(self):
        """Actions to be performed when the strategy is reset."""
        # Reset indicators here
        self.fast_ema.reset()
        self.slow_ema.reset()

Development

We aim to provide the most pleasant developer experience possible for this hybrid codebase of Python, Cython and Rust. Please refer to the Developer Guide for helpful information.

Contributing

Involvement from the trading community is a goal for this project, all help is welcome! Developers can open issues on GitHub to discuss proposed enhancements, changes, or to make bug reports. Questions and more general thoughts are best directed to a discussions thread.

Refer to the CONTRIBUTING.md for further information.

Please make all pull requests to the develop branch.

Community

Chat with contributors and active users of NautilusTrader on our Discord server! This is also the best place to monitor announcements, and learn about the latest features as they become available.

License

NautilusTrader is licensed under the GNU Lesser General Public License v3.0.

Contributors are also required to sign a standard Contributor License Agreement (CLA), which is administered automatically through CLA Assistant.


Copyright (C) 2015-2022 Nautech Systems Pty Ltd. All rights reserved. https://nautilustrader.io

nautechsystems

About

A high-performance algorithmic trading platform and event-driven backtester

Resources

License

Code of conduct

Security policy

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Python 56.3%
  • Cython 39.0%
  • Rust 4.2%
  • Other 0.5%