PFund: A Complete Algo-Trading Framework powered by Machine Learning and Data Engineering, TradFi, CeFi and DeFi ready.

Problem

Machine learning (AI) and data engineering (Big Data) fields are advancing every year, but everyday traders are not able to enjoy the benefits of these improvements, leading to a widening gap between retail traders and professional traders.

Solution

A modern algo-trading framework is needed to bridge the gap between algo-trading, machine learning and data engineering, empowering retail traders with state-of-the-art machine learning models and data engineering tools so that traders only need to focus on strategy research and the framework takes care of the rest.

---

PFund (/piː fʌnd/), which stands for "Personal Fund", is an algo-trading framework designed for using machine learning models natively to trade across TradFi (Traditional Finance, e.g. Interactive Brokers), CeFi (Centralized Finance, e.g. Binance) and DeFi (Decentralized Finance, e.g. dYdX), or in simple terms, Stocks and Cryptos.

Core Features

• Supports vectorized and event-driven backtesting with different resolutions of data, e.g. tick data, second data and minute data etc.
• Allows choosing your preferred data tool, e.g. pandas, polars, pyspark etc.
• Supports machine learning models, features, technical analysis indicators
• Trains machine learning models using your favorite frameworks, i.e. PFund is ML-framework agnostic
• Offers LEGO-style strategy and model building, allowing strategies to add other strategies, models to add other models
• Streamlines the algo-trading flow, from vectorized backtesting for strategy prototyping and event-driven backtesting for strategy development, to live trading for strategy deployment
• Enables parallel data processing, e.g. Interactive Brokers and Binance each have their own process for receiving data feeds
• Switches from backtesting to live trading by just changing ONE line of code!!
• Features a modern frontend using Mantine UI and TradingView's Charts library
• Supports manual/semi-manual trading via a trading app

As PFund is for trading only, for all the data workloads, there is a separate library to handle that:
PFeed - Data pipeline for algo-trading, helping traders in getting real-time and historical data, and storing them in a local data lake for quantitative research.

---

Table of Contents

• Project Status
• Installation
• Quick Start
  • Backtesting
  • Live Trading
  • Parameter Training / Hyperparameter Tuning
  • Building LEGO-Style Strategy and Model
• Model Hub
• Supported Trading Venues
• Related Projects
• Disclaimer

Project Status

Caution: PFund is at a VERY EARLY stage, use it at your own risk.

PFund is currently under active development, the framework design will be prioritized first over stability and scalability.

PFund for the time being only supports backtesting using Bybit and Yahoo Finance data for testing purpose.

Installation

Using Poetry (Recommended)

# [RECOMMENDED]: Trading + Backtesting + Machine Learning + Feature Engineering (e.g. feast, tsfresh, ta) + Analytics
poetry add "pfund[all]"

# [Trading + Backtesting + Machine Learning + Feature Engineering]:
poetry add "pfund[data,ml,fe]"

# [Trading + Backtesting + Machine Learning]:
poetry add "pfund[data,ml]"

# [Trading + Backtesting]:
poetry add "pfund[data]"

# [Trading only]:
poetry add pfund

# update to the latest version:
poetry update pfund

Using Pip

pip install pfund

# install the latest version:
pip install -U pfund

Checking your installation

$ pfund --version

Quick Start

Backtesting

import pfund as pf

# NOTE: for more exciting strategies, please visit pfund.ai
class YourStrategy(pf.Strategy):
    # triggered by bar/kline data (e.g. 1-minute data)
    def on_bar(self):
        # write your trading logic here
        pass


engine = pf.BacktestEngine(mode='vectorized')
strategy = engine.add_strategy(YourStrategy(), name='your_strategy')
strategy.add_data(
  'IB', 'AAPL', 'USD', 'STK', resolutions=['1d'],
  backtest={
    # NOTE: since IB does not provide any historical data for backtesting purpose, use data from 'YAHOO_FINANCE'
    'data_source': 'YAHOO_FINANCE',
    'start_date': '2024-01-01',
    'end_date': '2024-02-01',
  }
)
engine.run()

Live Trading

Just change one line of code, from 'BacktestEngine' to 'TradeEngine'. BOOM! you can now start live trading.

import pfund as pf

engine = pf.TradeEngine(env='LIVE')
strategy = engine.add_strategy(YourStrategy(), name='your_strategy')
strategy.add_data(
  'IB', 'AAPL', 'USD', 'STK', resolutions=['1d'],
  # for convenience, you can keep the kwarg `backtest`, `TradeEngine` will ignore it
  backtest={
    # NOTE: since IB does not provide any historical data for backtesting purpose, use data from 'YAHOO_FINANCE'
    'data_source': 'YAHOO_FINANCE',
    'start_date': '2024-01-01',
    'end_date': '2024-02-01',
  }
)
engine.run()

Parameter Training / Hyperparameter Tuning

The correct term should be "Hyperparameter Tuning", but since not all traders are familiar with machine learning, the framework uses a more well-known term "training".

import pfund as pf

engine = pf.TrainEngine()
strategy = engine.add_strategy(...)
strategy.add_data(...)
strategy.add_indicator(...)
engine.run()

Building LEGO-Style Strategy and Model

import pfund as pf

engine = pf.TradeEngine(env='LIVE')
strategy = engine.add_strategy(...)
strategy.add_data(...)
model = strategy.add_model(...)

model.add_data(...)  # using different data than strategy's
sub_model = model.add_model(...)  # YES, model can add another model to its use
# You can keep going: 
# sub_sub_model = sub_model.add_model(...)

engine.run()

Model Hub

Imagine a space where algo-traders can share their machine learning models with one another. Strategy and model development could be so much faster since you can build on top of an existing working model.

Model Hub is coming soon on PFund.ai, Stay Tuned!

---

Supported Trading Venues

Trading Venue	Vectorized Backtesting	Event-Driven Backtesting	Paper Trading	Live Trading
Bybit	🟢	🟡	🟡	🟡
*Interactive Brokers (IB)	🟡	🟡	🟡	🟡
Binance	🔴	🔴	🔴	🔴
OKX	🔴	🔴	🔴	🔴
*Alpaca	🔴	🔴	🔴	🔴
*Futu	🔴	🔴	🔴	🔴
dYdX	🔴	🔴	🔴	🔴

🟢 = finished
🟡 = in progress
🔴 = todo
* = use a separate data source (e.g. FirstRate Data) for backtesting

Related Projects

• PFeed — Data pipeline for algo-trading, helping traders in getting real-time and historical data, and storing them in a local data lake for quantitative research.
• PyTrade.org - A curated list of Python libraries and resources for algorithmic trading.

Disclaimer

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES, OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF, OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

This algo-trading framework is intended for educational and research purposes only. It should not be used for real trading without understanding the risks involved. Trading in financial markets involves significant risk, and there is always the potential for loss. Your trading results may vary. No representation is being made that any account will or is likely to achieve profits or losses similar to those discussed on this platform.

The developers of this framework are not responsible for any financial losses incurred from using this software. Users should conduct their due diligence and consult with a professional financial advisor before engaging in real trading activities.