AI-Powered Algorithmic Trading with Python
Algorithmic Trading in financial markets used to be the domain of big institutional players a few years ago. Nowadays, with the advent of newly built retail trading platforms, even individual traders with low funds available for trading can enter the Algorithmic Trading space with relatively low effort. Generally speaking, a typical notebook, a Python environment, several open-source packages, and an Internet connection are enough. You then only need to open a trading account — either a paper trading account to get started with, or a real trading account with say 100 USD/EUR/GBP initial capital.
This half-day training walks you through the steps necessary to start Algorithmic Trading using Python and additional open-source packages, such as NumPy, pandas, matplotlib, and scikit-learn. The idea is to get started by setting up a trading infrastructure first and using it to retrieve data, to formulate and backtest algorithmic trading strategies, to place orders, and to automate the algorithmic trading strategy for continuous deployment.
The trading platform used for this training is the one from Oanda for which you can open a demo account here https://www.oanda.com/eu-en/trading/.
Assuming that you already have a Python environment with the typical data science packages installed, the only additional packages required is https://github.com/yhilpisch/tpqoa. You can install it as follows:
pip install git+https://github.com/yhilpisch/tpqoa.gitIn order to make use of the package, you should create a credentials file, say with filename oanda.cfg, with the following structure:
[oanda]
account_id = XYZ-ABC-...
access_token = ZYXCAB...
account_type = practiceThe required information is found in the demo account that you have created. Once this is done, you can check whether you can connect to the Oanda API as follows:
Python 3.9.2 | packaged by conda-forge | (default, Feb 21 2021, 05:00:30)
Type 'copyright', 'credits' or 'license' for more information
IPython 7.22.0 -- An enhanced Interactive Python. Type '?' for help.In [1]: import tpqoaIn [2]: oanda = tpqoa.tpqoa('oanda.cfg')In [3]: data = oanda.get_history(instrument='EUR_USD',
...: start='2021-06-01',
...: end='2022-05-31',
...: granularity='D',
...: price='M')In [4]: data.head()
Out[4]:
o h l c volume complete
time
2021-05-31 21:00:00 1.22258 1.22545 1.22116 1.22138 81745 True
2021-06-01 21:00:00 1.22134 1.22266 1.21640 1.22104 84234 True
2021-06-02 21:00:00 1.22104 1.22142 1.21182 1.21271 95659 True
2021-06-03 21:00:00 1.21273 1.21856 1.21040 1.21668 93352 True
2021-06-06 21:00:00 1.21671 1.22018 1.21449 1.21894 64527 TrueIn [5]: data.tail()
Out[5]:
o h l c volume complete
time
2022-05-24 21:00:00 1.07362 1.07390 1.06421 1.06798 166318 True
2022-05-25 21:00:00 1.06819 1.07317 1.06626 1.07302 135007 True
2022-05-26 21:00:00 1.07257 1.07652 1.06968 1.07334 124909 True
2022-05-29 21:00:00 1.07375 1.07870 1.07262 1.07800 85924 True
2022-05-30 21:00:00 1.07775 1.07798 1.07278 1.07446 50284 FalseIn [6]:
With the API connection, you can also place orders easily:
In [6]: oanda.create_order('EUR_USD', units=1000)
{'id': '3142', 'time': '2022-05-31T08:22:26.949998325Z', 'userID': 13834683, 'accountID':
'101-004-13834683-001', 'batchID': '3141', 'requestID': '60980512881182334', 'type':
'ORDER_FILL', 'orderID': '3141', 'instrument': 'EUR_USD', 'units': '1000.0',
'gainQuoteHomeConversionFactor': '0.926305211066', 'lossQuoteHomeConversionFactor':
'0.935614811178', 'price': 1.07424, 'fullVWAP': 1.07424, 'fullPrice': {'type': 'PRICE',
'bids': [{'price': 1.07408, 'liquidity': '10000000'}], 'asks': [{'price': 1.07424, 'liquidity': '10000000'}],
'closeoutBid': 1.07408, 'closeoutAsk': 1.07424}, 'reason': 'MARKET_ORDER', 'pl':
'0.0', 'financing': '0.0', 'commission': '0.0', 'guaranteedExecutionFee': '0.0',
'accountBalance': '101359.1229', 'tradeOpened': {'tradeID': '3142', 'units': '1000.0',
'price': 1.07424, 'guaranteedExecutionFee': '0.0', 'halfSpreadCost': '0.0745',
'initialMarginRequired': '33.3'}, 'halfSpreadCost': '0.0745'}In [7]: oanda.create_order('EUR_USD', units=-1000) {'id': '3144', 'time': '2022-05-31T08:22:37.121812044Z', 'userID': 13834683, 'accountID':
'101-004-13834683-001', 'batchID': '3143', 'requestID': '60980512927332557', 'type':
'ORDER_FILL', 'orderID': '3143', 'instrument': 'EUR_USD', 'units': '-1000.0',
'gainQuoteHomeConversionFactor': '0.926279341195', 'lossQuoteHomeConversionFactor':
'0.935588681308', 'price': 1.07411, 'fullVWAP': 1.07411, 'fullPrice': {'type': 'PRICE',
'bids': [{'price': 1.07411, 'liquidity': '10000000'}], 'asks': [{'price': 1.07427,
'liquidity': '10000000'}], 'closeoutBid': 1.07411, 'closeoutAsk': 1.07427}, 'reason':
'MARKET_ORDER', 'pl': '-0.1216', 'financing': '0.0', 'commission': '0.0',
'guaranteedExecutionFee': '0.0', 'accountBalance': '101359.0013', 'tradesClosed':
[{'tradeID': '3142', 'units': '-1000.0', 'price': 1.07411, 'realizedPL': '-0.1216',
'financing': '0.0', 'guaranteedExecutionFee': '0.0', 'halfSpreadCost': '0.0745'}], 'halfSpreadCost': '0.0745'}In [8]:
Real-time, or streaming, data is also at your fingertips:
In [8]: oanda.stream_data('EUR_USD', stop=10)
2022-05-31T08:24:05.263842391Z 1.07406 1.07422
2022-05-31T08:24:07.321639227Z 1.07405 1.0742
2022-05-31T08:24:08.445782000Z 1.07405 1.07421
2022-05-31T08:24:09.585760607Z 1.07405 1.07419
2022-05-31T08:24:10.336223320Z 1.07406 1.07419
2022-05-31T08:24:11.235281237Z 1.07409 1.07423
2022-05-31T08:24:11.397952230Z 1.07406 1.0742
2022-05-31T08:24:11.572900914Z 1.07405 1.0742
2022-05-31T08:24:12.000995719Z 1.07405 1.07419
2022-05-31T08:24:14.192202399Z 1.07404 1.0742In [9]:
These are already the basic ingredients required to get started with Algorithmic Trading: historical data, the ability to place orders and the ability to retrieve data in real-time. In the training, we will make use of these capabilities to implement an algorithmic trading system in a basic fashion. In particular, we will make use of AI (= machine learning) to come up with the algorithmic trading strategy, representing the core of the system.
Read more data science articles on OpenDataScience.com, including tutorials and guides from beginner to advanced levels! Subscribe to our weekly newsletter here and receive the latest news every Thursday. You can also get data science training on-demand wherever you are with our Ai+ Training platform. Subscribe to our fast-growing Medium Publication too, the ODSC Journal, and inquire about becoming a writer.
