Zipline tutorial¶

In [1]:

# Enable plotting
%pylab --no-import-all inline

# Provides tabular data structure and much more!
import pandas as pd

import datetime
figsize(10,5)

Populating the interactive namespace from numpy and matplotlib

In [3]:

from zipline.data import load_from_yahoo

data = load_from_yahoo(stocks=['AAPL'],
                       start=datetime.datetime(2006, 1, 1))

AAPL

In [27]:

data.head()

Out[27]:

	AAPL
Date
2006-01-03 00:00:00+00:00	72.28
2006-01-04 00:00:00+00:00	72.49
2006-01-05 00:00:00+00:00	71.92
2006-01-06 00:00:00+00:00	73.78
2006-01-09 00:00:00+00:00	73.53

5 rows × 1 columns

In [16]:

data['AAPL'].plot()

Out[16]:

<matplotlib.axes.AxesSubplot at 0xf575910>

Simplest algorithm possible: Always buy apple¶

In [7]:

from zipline.algorithm import TradingAlgorithm
from zipline.api import order

def initialize(context):
    print('Initialized!')
    context.last_price = 0

def handle_data(context, data):
    stock = 'AAPL'
    if context.last_price < data[stock].price:
        order(stock, 10)
    else:
        order(stock, -10)
    
    context.last_price = data[stock].price

In [8]:

# Instantiate our algorithm
my_algo = TradingAlgorithm(initialize=initialize,
                           handle_data=handle_data)

# Run algorithm on data
perf_buy_apple = my_algo.run(data)

[2014-01-16 12:33] INFO: Performance: Simulated 2022 trading days out of 2022.
[2014-01-16 12:33] INFO: Performance: first open: 2006-01-03 14:31:00+00:00
[2014-01-16 12:33] INFO: Performance: last close: 2014-01-14 21:00:00+00:00

Initialized!

In [9]:

perf_buy_apple.head()

Out[9]:

	capital_used	ending_cash	ending_value	orders	period_close	period_open	pnl	portfolio_value	positions	returns	starting_cash	starting_value	transactions
2006-01-03 21:00:00	0.000000	100000.000000	0.0	[{u'status': 0, u'created': 2006-01-03 00:00:0...	2006-01-03 21:00:00+00:00	2006-01-03 14:31:00+00:00	0.000000	100000.000000	[]	0.000000	100000.000000	0.0	[]
2006-01-04 21:00:00	-725.207249	99274.792751	724.9	[{u'status': 1, u'created': 2006-01-03 00:00:0...	2006-01-04 21:00:00+00:00	2006-01-04 14:31:00+00:00	-0.307249	99999.692751	[{u'amount': 10, u'last_sale_price': 72.49, u'...	-0.000003	100000.000000	0.0	[{u'commission': 0.3, u'amount': 10, u'sid': u...
2006-01-05 21:00:00	-719.507192	98555.285559	1438.4	[{u'status': 1, u'created': 2006-01-04 00:00:0...	2006-01-05 21:00:00+00:00	2006-01-05 14:31:00+00:00	-6.007192	99993.685559	[{u'amount': 20, u'last_sale_price': 71.92, u'...	-0.000060	99274.792751	724.9	[{u'commission': 0.3, u'amount': 10, u'sid': u...
2006-01-06 21:00:00	737.492622	99292.778181	737.8	[{u'status': 1, u'created': 2006-01-05 00:00:0...	2006-01-06 21:00:00+00:00	2006-01-06 14:31:00+00:00	36.892622	100030.578181	[{u'amount': 10, u'last_sale_price': 73.78, u'...	0.000369	98555.285559	1438.4	[{u'commission': 0.3, u'amount': -10, u'sid': ...
2006-01-09 21:00:00	-735.607353	98557.170828	1470.6	[{u'status': 1, u'created': 2006-01-06 00:00:0...	2006-01-09 21:00:00+00:00	2006-01-09 14:31:00+00:00	-2.807353	100027.770828	[{u'amount': 20, u'last_sale_price': 73.53, u'...	-0.000028	99292.778181	737.8	[{u'commission': 0.3, u'amount': 10, u'sid': u...

5 rows × 13 columns

In [10]:

perf_buy_apple.portfolio_value.plot()

Out[10]:

<matplotlib.axes.AxesSubplot at 0x4109a50>

A classic: Dual Moving Average crossover strategy.¶

In [21]:

from zipline.transforms import MovingAverage
from zipline.api import order_target, record

def initialize_dma(context):
    # Add 2 mavg transforms, one with a long window, one
    # with a short window.
    # Note that this is bound to change soon and will be easier.
    context.add_transform(MovingAverage, 'short_mavg', ['price'],
                          window_length=100)

    context.add_transform(MovingAverage, 'long_mavg', ['price'],
                          window_length=300)

def handle_data_dma(context, data):
    short_mavg = data['AAPL'].short_mavg['price']
    long_mavg = data['AAPL'].long_mavg['price']

    if short_mavg > long_mavg:
        order_target('AAPL', 100)
    elif short_mavg < long_mavg:
        order_target('AAPL', 0)
        
    record(short_mavg=short_mavg,
           long_mavg=long_mavg)

In [23]:

dma = TradingAlgorithm(initialize=initialize_dma,
                       handle_data=handle_data_dma)
perf_dma = dma.run(data)

[2014-01-15 22:02] INFO: Performance: Simulated 2022 trading days out of 2022.
[2014-01-15 22:02] INFO: Performance: first open: 2006-01-03 14:31:00+00:00
[2014-01-15 22:02] INFO: Performance: last close: 2014-01-14 21:00:00+00:00

In [34]:

perf_dma.long_mavg

Out[34]:

2006-01-03 21:00:00    72.280000
2006-01-04 21:00:00    72.385000
2006-01-05 21:00:00    72.230000
2006-01-06 21:00:00    72.617500
2006-01-09 21:00:00    72.800000
2006-01-10 21:00:00    73.696667
2006-01-11 21:00:00    74.757143
2006-01-12 21:00:00    75.600000
2006-01-13 21:00:00    76.395556
2006-01-17 21:00:00    76.947000
2006-01-18 21:00:00    77.202727
2006-01-19 21:00:00    77.137500
2006-01-20 21:00:00    76.863077
2006-01-23 21:00:00    76.737143
2006-01-24 21:00:00    76.522667
...
2013-12-23 21:00:00    480.488300
2013-12-24 18:00:00    480.342233
2013-12-26 21:00:00    480.167267
2013-12-27 21:00:00    479.930967
2013-12-30 21:00:00    479.692867
2013-12-31 21:00:00    479.515167
2014-01-02 21:00:00    479.384967
2014-01-03 21:00:00    479.135967
2014-01-06 21:00:00    478.963700
2014-01-07 21:00:00    478.767233
2014-01-08 21:00:00    478.605767
2014-01-09 21:00:00    478.439100
2014-01-10 21:00:00    478.288600
2014-01-13 21:00:00    478.143433
2014-01-14 21:00:00    478.097700
Name: long_mavg, Length: 2022

In [24]:

fig = plt.figure()
ax1 = fig.add_subplot(211)
perf_dma.portfolio_value.plot(ax=ax1)

ax2 = fig.add_subplot(212)
data['AAPL'].plot(ax=ax2)
perf_dma[['short_mavg', 'long_mavg']].plot(ax=ax2)

perf_dma_trans = perf_dma.ix[[t != [] for t in perf_dma.transactions]]
buys = perf_dma_trans.ix[[t[0]['amount'] > 0 for t in perf_dma_trans.transactions]]
sells = perf_dma_trans.ix[[t[0]['amount'] < 0 for t in perf_dma_trans.transactions]]
ax2.plot(buys.index, perf_dma.short_mavg.ix[buys.index],
         '^', markersize=10, color='m')
ax2.plot(sells.index, perf_dma.short_mavg.ix[sells.index],
         'v', markersize=10, color='k')
plt.legend(loc=0)

Out[24]:

<matplotlib.legend.Legend at 0xc3b4850>