zipline-tej 0.0.40

A Pythonic backtester for trading algorithms

Installation

Used packages and environment

• Main package: Zipline-reloaded 2.2.0
• Python 3.8
• Microsoft Windows OS or macOS
• Other Python dependency packages: Pandas, Numpy, Logbook, Exchange-calendars

How to install Zipline Reloaded modified by TEJ

• We're going to illustrate under anaconda environment, so we suggest using Anaconda as development environment.

• Download dependency packages.

1. Windows (zipline-tej.yml)

2. Mac (zipline-tej_mac.yml)

• Start an Anaconda (base) prompt, create an virtual environment and install the appropriate versions of packages: (We strongly recommand using virtual environment to keep every project independent.) (reason)

Windows Users
# change directionary to the folder exists zipline-tej.yml
$ cd <C:\Users\username\Downloads>

# create virtual env
$ conda env create -f zipline-tej.yml

# activate virtual env
$ conda activate zipline-tej
Mac Users
# change directionary to the folder exists zipline-tej_mac.yml
$ cd <C:\Users\username\Downloads>

# create virtual env
$ conda env create -f zipline-tej_mac.yml

# activate virtual env
$ conda activate zipline-tej

Also, if you are familiar with Python enough, you can create a virtual environment without zipline-tej.yml and here's the sample :

# create virtual env
$ conda create -n <env_name> python=3.8

# activate virtual env
$ conda activate <env_name>

# download dependency packages
$ conda install -c conda-forge -y ta-lib
$ conda install -c conda-forge nb_conda_kernels
$ conda install -y notebook=6.4.11
$ conda install -y xlrd=2.0.1
$ conda install -y openpyxl=3.0.9
$ pip install zipline-tej

• Notice that we need to download TA-lib at first, so that we can download zipline-tej successfully.

Exchange Calendar Issues

We're now developing specfic on Taiwan securities backtesting strategy, so we're using the unique trading calendar created by ourselves. download

After downloaded the calendar file above, overwrite calendar_utils.py in exchange_calendars folder and add exchange_calendar_tejxtai.py.

* Navigate to the exchange_calendars folder within site packages. This is typically located at

C:\Users\username\Anaconda3\envs\zipline-tej\Lib\site-packages\exchange_calendars

Mac may located at

C:\Users\username\Anaconda3\envs\zipline-tej\Lib\python3.8\site-packages\exchange_calendars

But some Windows users may located at

C:\Users\username\AppData\Roaming\Python\Python38\Scripts

which we aren't pleased to see. So if this happened, we suggest to put exchange-calendars folder to former path we mentioned above.

Quick start

CLI Interface

The following code implements a simple buy_and_hold trading algorithm.

from zipline.api import order, record, symbol

def initialize(context):
    context.asset = symbol("2330")
    
def handle_data(context, data):
    order(context.asset, 10)
    record(TSMC=data.current(context.asset, "price"))
    
def analyze(context=None, results=None):
    import matplotlib.pyplot as plt

    # Plot the portfolio and asset data.
    ax1 = plt.subplot(211)
    results.portfolio_value.plot(ax=ax1)
    ax1.set_ylabel("Portfolio value (TWD)")
    ax2 = plt.subplot(212, sharex=ax1)
    results.TSMC.plot(ax=ax2)
    ax2.set_ylabel("TSMC price (TWD)")

    # Show the plot.
    plt.gcf().set_size_inches(18, 8)
    plt.show()

You can then run this algorithm using the Zipline CLI. But first, you need to download some market data with historical prices and trading volumes:

• Before ingesting data, you have to set some environment variables as follow:

# setting TEJAPI_KEY to get permissions loading data
$ set TEJAPI_KEY=<your_key>

# setting download ticker
$ set ticker=2330 2317

# setting backtest period
$ set mdate=20200101 20220101

• Ingest and run backtesting algorithm

$ zipline ingest -b tquant
$ zipline run -f buy_and_hold.py  --start 20200101 --end 20220101 -o bah.pickle --no-benchmark --trading-calendar TEJ_XTAI

Then, the resulting performance DataFrame is saved as bah.pickle, which you can load and analyze from Python.

Jupyter Notebook

Change Anaconda kernel

• Since we've downloaded package "nb_conda_kernels", we should be able to change kernel in jupyter notebook.

How to new a notebook using specific kernel

(1) Open anaconda prompt

(2) Enter the command as follow :

# First one can be ignore if already in environment of zipline-tej
$ conda activate zipline-tej 
# start a jupyter notebook
$ jupyter notebook 

(3) Start a notebook and select Python[conda env:zipline-tej]

(4)(Optional) If you have already written a notebook, you can open it and change kernel by clicking the "Kernel" in menu and "Change kernel" to select the specfic kernel.

Set environment variables TEJAPI_KEY, ticker and mdate

* ticker would be your target ticker symbol, and it should be a string. If there're more than one ticker needed, use " ", "," or ";" to split them apart.

* mdate refers the begin date and end date, use " ", "," or ";" to split them apart.

In[1]:
import os    
os.environ['TEJAPI_KEY'] = <your_key>    
os.environ['ticker'] ='2330 2317'     
os.environ['mdate'] ='20200101 20220101'  

Call ingest to download data to ~\.zipline

In[2]:    
!zipline ingest -b tquant
[Out]: 
Merging daily equity files:
[YYYY-MM-DD HH:mm:ss.ssssss] INFO: zipline.data.bundles.core: Ingesting tquant.

Design the backtesting strategy

In[3]:
from zipline.api import order, record, symbol

def initialize(context):
    context.asset = symbol("2330")
    
def handle_data(context, data):
    order(context.asset, 10)
    record(TSMC=data.current(context.asset, "price"))
    
def analyze(context=None, results=None):
    import matplotlib.pyplot as plt

    # Plot the portfolio and asset data.
    ax1 = plt.subplot(211)
    results.portfolio_value.plot(ax=ax1)
    ax1.set_ylabel("Portfolio value (TWD)")
    ax2 = plt.subplot(212, sharex=ax1)
    results.TSMC.plot(ax=ax2)
    ax2.set_ylabel("TSMC price (TWD)")

    # Show the plot.
    plt.gcf().set_size_inches(18, 8)
    plt.show()

Run backtesting algorithm and plot

In[4]:
from zipline import run_algorithm
import pandas as pd
from zipline.utils.calendar_utils import get_calendar
trading_calendar = get_calendar('TEJ_XTAI')

start = pd.Timestamp('20200103', tz ='utc' )
end = pd.Timestamp('20211230', tz='utc')

result = run_algorithm(start=start,
                  end=end,
                  initialize=initialize,
                  capital_base=1000000,
                  handle_data=handle_data,
                  bundle='tquant',
                  trading_calendar=trading_calendar,
                  analyze=analyze,
                  data_frequency='daily'
                  )
[Out]:

Show trading process

In[5]: 
result
[Out]:

	period_open	period_close	starting_value	ending_value	starting_cash	ending_cash	portfolio_value	longs_count	shorts_count	long_value	...	treasury_period_return	trading_days	period_label	algo_volatility	benchmark_period_return	benchmark_volatility	algorithm_period_return	alpha	beta	sharpe
2020-01-03 05:30:00+00:00	2020-01-03 01:01:00+00:00	2020-01-03 05:30:00+00:00	0.0	0.0	1.000000e+06	1.000000e+06	1.000000e+06	0	0	0.0	...	0.0	1	2020-01	NaN	0.0	NaN	0.000000	None	None	NaN
2020-01-06 05:30:00+00:00	2020-01-06 01:01:00+00:00	2020-01-06 05:30:00+00:00	0.0	3320.0	1.000000e+06	9.966783e+05	9.999983e+05	1	0	3320.0	...	0.0	2	2020-01	0.000019	0.0	0.0	-0.000002	None	None	-11.224972
2020-01-07 05:30:00+00:00	2020-01-07 01:01:00+00:00	2020-01-07 05:30:00+00:00	3320.0	6590.0	9.966783e+05	9.933817e+05	9.999717e+05	1	0	6590.0	...	0.0	3	2020-01	0.000237	0.0	0.0	-0.000028	None	None	-10.038514
2020-01-08 05:30:00+00:00	2020-01-08 01:01:00+00:00	2020-01-08 05:30:00+00:00	6590.0	9885.0	9.933817e+05	9.900850e+05	9.999700e+05	1	0	9885.0	...	0.0	4	2020-01	0.000203	0.0	0.0	-0.000030	None	None	-9.298128
2020-01-09 05:30:00+00:00	2020-01-09 01:01:00+00:00	2020-01-09 05:30:00+00:00	9885.0	13500.0	9.900850e+05	9.867083e+05	1.000208e+06	1	0	13500.0	...	0.0	5	2020-01	0.001754	0.0	0.0	0.000208	None	None	5.986418
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
2021-12-24 05:30:00+00:00	2021-12-24 01:01:00+00:00	2021-12-24 05:30:00+00:00	2920920.0	2917320.0	-1.308854e+06	-1.314897e+06	1.602423e+06	1	0	2917320.0	...	0.0	484	2021-12	0.232791	0.0	0.0	0.602423	None	None	1.170743
2021-12-27 05:30:00+00:00	2021-12-27 01:01:00+00:00	2021-12-27 05:30:00+00:00	2917320.0	2933040.0	-1.314897e+06	-1.320960e+06	1.612080e+06	1	0	2933040.0	...	0.0	485	2021-12	0.232577	0.0	0.0	0.612080	None	None	1.182864
2021-12-28 05:30:00+00:00	2021-12-28 01:01:00+00:00	2021-12-28 05:30:00+00:00	2933040.0	2982750.0	-1.320960e+06	-1.327113e+06	1.655637e+06	1	0	2982750.0	...	0.0	486	2021-12	0.233086	0.0	0.0	0.655637	None	None	1.237958
2021-12-29 05:30:00+00:00	2021-12-29 01:01:00+00:00	2021-12-29 05:30:00+00:00	2982750.0	2993760.0	-1.327113e+06	-1.333276e+06	1.660484e+06	1	0	2993760.0	...	0.0	487	2021-12	0.232850	0.0	0.0	0.660484	None	None	1.243176
2021-12-30 05:30:00+00:00	2021-12-30 01:01:00+00:00	2021-12-30 05:30:00+00:00	2993760.0	2995050.0	-1.333276e+06	-1.339430e+06	1.655620e+06	1	0	2995050.0	...	0.0	488	2021-12	0.232629	0.0	0.0	0.655620	None	None	1.235305

488 rows × 38 columns

Common errors

• NotSessionError : The date of algorithm start date or end date is not available in trading algorithm.
  • Solution : Adjust start date or end date to align trading calendar.
• DateOutOfBounds : The trading calendar would update every day, but it would be fixed on the FIRST TIME executed date in Jupyter Notebook.
  • Solution : Restart Jupyter Notebook kernel.

More Zipline Tutorials

• For more tutorials

Suggestions

• To get TEJAPI_KEY (link)
• TEJ Official Website