RiskEngine: Iteration 2

Branch	Version	Status
`master`
`nightly`
`develop`

Platform	Rust	Python
`Linux (x86_64)`	1.78.0+	3.10+
`macOS (x86_64)`	1.78.0+	3.10+
`macOS (arm64)`	1.78.0+	3.10+
`Windows (x86_64)`	1.78.0+	3.10+

Website: https://nautilustrader.io
Docs: https://docs.nautilustrader.io
Support: [email protected]

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 - Core written in Rust with asynchronous networking using tokio
Reliable - Type safety and thread safety through Rust. 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)

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 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 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 and PyO3, 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.

This project makes the Soundness Pledge:

“The intent of this project is to be free of soundness bugs. The developers will do their best to avoid them, and welcome help in analyzing and fixing them.”

Architecture (data flow)

Integrations

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

Name	ID	Type	Docs
Betfair	`BETFAIR`	Sports Betting Exchange	Guide
Binance	`BINANCE`	Crypto Exchange (CEX)	Guide
Binance US	`BINANCE`	Crypto Exchange (CEX)	Guide
Binance Futures	`BINANCE`	Crypto Exchange (CEX)	Guide
Bybit	`BYBIT`	Crypto Exchange (CEX)	Guide
Databento	`DATABENTO`	Data Provider	Guide
Interactive Brokers	`INTERACTIVE_BROKERS`	Brokerage (multi-venue)	Guide

ID: The default client ID for the integrations adapter clients
Type: The type of integration (often the venue type)

Status

building - Under construction and likely not in a usable state
beta - Completed to a minimally working state and in a 'beta' testing phase
stable - Stabilized feature set and API, the integration has been tested by both developers and users to a reasonable level (some bugs may still remain)

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

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.

For MacBook Pro M1/M2, make sure your Python installed using pyenv is configured with --enable-shared:

PYTHON_CONFIGURE_OPTS="--enable-shared" pyenv install <python_version>

See https://pyo3.rs/latest/getting_started#virtualenvs.

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.

Install rustup (the Rust toolchain installer):
- Linux and macOS:
```
curl https://sh.rustup.rs -sSf | sh
```
- Windows:
  - Download and install rustup-init.exe
  - Install "Desktop development with C++" with Build Tools for Visual Studio 2019
Enable cargo in the current shell:
- Linux and macOS:
```
source $HOME/.cargo/env
```
- Windows:
  - Start a new PowerShell
Install poetry (or follow the installation guide on their site):

curl -sSL https://install.python-poetry.org | python3 -
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 --only main --all-extras

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
nightly branch may contain experimental features and is generally merged from develop branch daily, and also when required
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 for development. It provides the following targets:

make install -- Installs in release build mode with main, dev and test dependencies then installs the package using poetry (default)
make install-debug -- Same as make install but with debug build mode
make install-just-deps -- Installs just the main, dev and test dependencies (does not install package)
make install-just-deps-all -- Same as make install-just-deps and additionally installs docs dependencies
make build -- Runs the build script in release build mode (default)
make build-debug -- Runs the build script in debug build mode
make build-wheel -- Runs the Poetry build with a wheel format in release mode
make build-wheel-debug -- Runs the Poetry build with a wheel format in debug mode
make clean -- CAUTION Cleans all non-source artifacts from the repository
make docs -- Builds the documentation HTML using Sphinx
make pre-commit -- Runs the pre-commit checks over all files
make ruff -- Runs ruff over all files using the pyproject.toml config
make outdated -- Runs commands to show outdated dependencies for both Rust and Python
make pytest -- Runs all tests with pytest (except performance tests)
make pytest-coverage -- Same as make pytest and additionally runs with test coverage and produces a report

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:3.11-slim with the following image variant tags:

nautilus_trader:latest has the latest release version installed
nautilus_trader:nightly has the head of the nightly branch installed
jupyterlab:latest has the latest release version installed along with jupyterlab and an example backtest notebook with accompanying data
jupyterlab:nightly has the head of the nightly 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> --platform linux/amd64

You can launch the backtest example container by running:

docker pull ghcr.io/nautechsystems/jupyterlab:nightly --platform linux/amd64
docker run -p 8888:8888 ghcr.io/nautechsystems/jupyterlab:nightly

Then open your browser at the following address:

http://127.0.0.1:8888/lab

⚠️ WARNING

NautilusTrader currently exceeds the rate limit for Jupyter notebook logging (stdout output), this is why log_level in the examples is set to ERROR. If you lower this level to see more logging then the notebook will hang during cell execution. A fix is currently being investigated which involves either raising the configured rate limits for Jupyter, or throttling the log flushing from Nautilus. jupyterlab/jupyterlab#12845 https://github.com/deshaw/jupyterlab-limit-output

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) -> None:
        super().__init__(config)

        # Configuration
        self.instrument_id = config.instrument_id
        self.bar_type = 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: Instrument | None = None  # Initialized in on_start

    def on_start(self) -> None:
        """
        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) -> None:
        """
        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) -> None:
        """
        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) -> None:
        """
        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) -> None:
        """
        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) -> None:
        """
        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. Refer to the Developer Guide for helpful information.

cargo-nextest is the standard Rust test runner for NautilusTrader. You can install it by running:

cargo install cargo-nextest

Contributing

Thank you for considering contributing to Nautilus Trader! We welcome any and all help to improve the project. If you have an idea for an enhancement or a bug fix, the first step is to open an issue on GitHub to discuss it with the team. This helps to ensure that your contribution will be well-aligned with the goals of the project and avoids duplication of effort.

Once you're ready to start working on your contribution, make sure to follow the guidelines outlined in the CONTRIBUTING.md file. This includes signing a Contributor License Agreement (CLA) to ensure that your contributions can be included in the project.

Note that all pull requests should be made to the develop branch. This is where new features and improvements are integrated before being released to the public.

Thank you again for your interest in Nautilus Trader! We look forward to reviewing your contributions and working with you to improve the project.

Community

Join our community of users and contributors on Discord to chat and stay up-to-date with the latest announcements and features of NautilusTrader. Whether you're a developer looking to contribute or just want to learn more about the platform, all are welcome on our server.

License

The source code for NautilusTrader is available on GitHub under the GNU Lesser General Public License v3.0. Contributions to the project are welcome and require the completion of a standard Contributor License Agreement (CLA).

NautilusTrader is developed and maintained by Nautech Systems, a technology company specializing in the development of high-performance trading systems. Although the project utilizes the Rust programming language and benefits from its ecosystem, Nautech Systems is not affiliated with the Rust Foundation, and this project is not an official work of the Rust Foundation. For more information, visit https://nautilustrader.io.

	df_ticks_o = df_ticks_o[(df_ticks_h[["bid_size"]] > 0).all(axis=1)]
	df_ticks_h = df_ticks_h[(df_ticks_h[["bid_size"]] > 0).all(axis=1)]
	df_ticks_l = df_ticks_l[(df_ticks_l[["bid_size"]] > 0).all(axis=1)]
	df_ticks_c = df_ticks_c[(df_ticks_c[["bid_size"]] > 0).all(axis=1)]

	df_ticks_o = df_ticks_o[(df_ticks_h[["ask_size"]] > 0).all(axis=1)]
	df_ticks_h = df_ticks_h[(df_ticks_h[["ask_size"]] > 0).all(axis=1)]
	df_ticks_l = df_ticks_l[(df_ticks_l[["ask_size"]] > 0).all(axis=1)]
	df_ticks_c = df_ticks_c[(df_ticks_c[["ask_size"]] > 0).all(axis=1)]

N samples	np_mean	fast_mean	fast_mean_new
5	10.6us (11.9)	0.7us (0.8)	0.9us (1.0)
10	12.0us (13.3)	1.4us (1.5)	0.9us (1.0)
20	11.8us (12.9)	2.5us (2.8)	0.9us (1.0)
50	11.3us (14.4)	4.7us (5.9)	0.8us (1.0)
100	11.0us (11.0)	8.7us (8.7)	1.0us (1.0)
200	11.9us (11.7)	17.7us (17.4)	1.0us (1.0)
500	10.5us (7.6)	42.0us (30.4)	1.4us (1.0)
1000	11.4us (6.3)	82.7us (45.6)	1.8us (1.0)
2000	12.4us (4.4)	162.5us (57.4)	2.8us (1.0)
5000	13.1us (2.3)	449.3us (79.3)	5.7us (1.0)
10000	14.2us (1.4)	837.3us (83.6)	10.0us (1.0)
20000	17.2us (0.9)	1675.0us (84.4)	19.8us (1.0)
50000	28.2us (0.6)	4182.1us (92.8)	45.1us (1.0)
100000	42.7us (0.5)	8287.9us (94.3)	87.9us (1.0)
1000000	816.3us (0.9)	84214.4us (93.2)	903.9us (1.0)
10000000	10978.4us (1.2)	882933.4us (96.8)	9124.7us (1.0)

N samples	np_std	fast_std	fast_std_new
5	33.6us (25.4)	1.8us (1.3)	1.3us (1.0)
10	37.2us (27.9)	3.4us (2.5)	1.3us (1.0)
20	34.1us (30.1)	6.8us (6.0)	1.1us (1.0)
50	34.2us (29.4)	13.4us (11.5)	1.2us (1.0)
100	33.9us (24.7)	26.3us (19.1)	1.4us (1.0)
200	37.1us (21.8)	51.9us (30.6)	1.7us (1.0)
500	34.9us (15.7)	132.8us (59.9)	2.2us (1.0)
1000	37.6us (12.7)	254.9us (86.2)	3.0us (1.0)
2000	42.5us (8.1)	531.7us (102.0)	5.2us (1.0)
5000	48.5us (4.7)	1336.0us (129.5)	10.3us (1.0)
10000	57.6us (2.8)	2630.5us (126.6)	20.8us (1.0)
20000	67.6us (1.8)	5192.6us (141.0)	36.8us (1.0)
50000	100.4us (1.1)	13126.2us (147.7)	88.9us (1.0)
100000	149.8us (0.8)	27235.6us (151.4)	179.8us (1.0)
1000000	4524.3us (2.5)	267034.9us (147.7)	1807.9us (1.0)
10000000	50570.1us (2.8)	2748350.5us (151.2)	18173.4us (1.0)

nautechsystems / nautilus_trader Goto Github PK

nautilus_trader's Introduction

Introduction

Features

Why NautilusTrader?

Why Python?

What is Rust?

Architecture (data flow)

Integrations

Status

Installation

From PyPI

From Source

Versioning and releases

Branches

Makefile

Examples

Docker

Minimal Strategy

Development

Contributing

Community

License

nautilus_trader's People

Contributors

Stargazers

Watchers

Forkers

nautilus_trader's Issues

Windows support (continue #175)

Starting points - platforms, libraries and nautilus_trader versions used:

Windows support problems and solutions:

A) Low-level code compatibility problems (There are only 3)

B) External libraries and tools compatibility 'problems'

Testing:

Conclusion:

Using Cython compiler directives to detect bugs/improvements/corrections

(1) - fast_mean, fast_std

Table 1 (comparison of 'mean' сomputation with different methods)

Table 2 (comparison of 'std' сomputation with different methods)

Conclusion

Background

Considerations

API changes

Questions

Recommend Projects

Recommend Topics

Recommend Org

Jobs