options_backtester/backtester/examples/main_example.ipynb at a4c728cc63226702b5d4f2f8ca43913d182c9ec2

In [1]:

import os
import sys
BACKTESTER_DIR = os.path.realpath(os.path.join(os.getcwd(), '..', '..'))
DATA_DIR = os.path.join(BACKTESTER_DIR, 'data')
OPTIONS_DATA = os.path.join(DATA_DIR, 'SPX_2017.csv')
STOCKS_DATA = os.path.join(DATA_DIR, 'portfolio_data_2017.csv')
sys.path.append(BACKTESTER_DIR) # Add backtester base dir to $PYTHONPATH

In [2]:

import pyfolio as pf
import matplotlib.pyplot as plt

from backtester import Backtest
from backtester.strategy import Strategy, StrategyLeg
from backtester.enums import Type, Direction, Stock
from backtester.datahandler import HistoricalOptionsData, TiingoData, Schema
from backtester.statistics import monthly_returns_heatmap, returns_histogram, returns_chart

/Users/jrchatruc/anaconda3/lib/python3.7/site-packages/pyfolio/pos.py:27: UserWarning: Module "zipline.assets" not found; mutltipliers will not be applied to position notionals.
  'Module "zipline.assets" not found; mutltipliers will not be applied' +

In [3]:

plt.style.use("seaborn")
plt.rcParams["figure.figsize"] = (14, 8)

As an example, we run a 2017 backtest of a portfolio consisting of a 97% allocation in stocks and 3% in options. The backtest will run for the time period spanned by the data, performing periodic rebalances with a user given frequency.

Stock data is taken from Tiingo.

In [ ]:

%env TIINGO_API_KEY=your_tiingo_api_key

In [ ]:

api_key = os.environ["TIINGO_API_KEY"]

start = datetime.datetime(2017, 1, 1)
end = datetime.datetime(2017, 12, 31)
tickers = ["VOO", "TUR", "RSX", "EWY", "EWS", "VTIP", "TLT", "BWX", "PDBC", "IAU", "VNQI"]

symbols = pdr.get_data_tiingo(tickers, api_key=api_key, start=start, end=end)

In [ ]:

save_path = os.path.join(DATA_DIR, 'portfolio_data_2017.csv')
symbols.to_csv(save_path)

Use TiingoData to load your stocks csv. Data must include date, adjClose and symbol columns to work.

In [4]:

stock_data = TiingoData(STOCKS_DATA)

In [5]:

stock_data.head()

Out [5]:

	symbol	date	close	high	low	open	volume	adjClose	adjHigh	adjLow	adjOpen	adjVolume	splitFactor
0	VOO	2017-01-03	206.74	207.33	205.560	206.68	4750181	194.997818	195.554308	193.884838	194.941226	4750181	1.0
1	VOO	2017-01-04	207.96	208.18	207.120	207.20	4622614	196.148526	196.356030	195.356235	195.431691	4622614	1.0
2	VOO	2017-01-05	207.80	208.04	207.013	207.75	2772065	195.997613	196.223982	195.255312	195.950453	2772065	1.0
3	VOO	2017-01-06	208.61	209.09	207.400	207.99	2194576	196.761608	197.214345	195.620332	196.176822	2194576	1.0
4	VOO	2017-01-09	207.95	208.48	207.885	208.34	1705181	196.139094	196.638991	196.077785	196.506943	1705181	1.0

To create a stock portfolio, use Stock as Stock(name, percentage), where name should match the name given in the symbol column and percentage is the percentage (from 0 to 1) allocated to that specific asset. Afterwards, store them in a list.

In [6]:

VOO = Stock('VOO', 0.1)
TUR = Stock('TUR', 0.05)
RSX = Stock('RSX', 0.05)
EWY = Stock('EWY', 0.05)
EWS = Stock('EWS', 0.05)
VTIP = Stock('VTIP', 0.10)
TLT = Stock('TLT', 0.20)
BWX = Stock('BWX', 0.10)
PDBC = Stock('PDBC', 0.05)
IAU = Stock('IAU', 0.15)
VNQI = Stock('VNQI', 0.10)

Make sure the sum of the percentages is 1.

In [7]:

0.10 + 0.05 + 0.05 + 0.05 + 0.05 + 0.10 + 0.20 + 0.10 + 0.05 + 0.15 + 0.10

Out [7]:

1.0000000000000002

In [8]:

stocks = [
    VOO,
    TUR,
    RSX,
    EWY,
    EWS,
    VTIP,
    TLT,
    BWX,
    PDBC,
    IAU,
    VNQI
]

To load your options data, use HistoricalOptionsData. Both this and Stock make use of an underlying schema to map the expected column names to the ones in your data. Schema.stocks() and Schema.options() will return the default mappings (i.e, the ones that don't change anything). If you need to change some values in them use update like so:

In [9]:

options_schema = Schema.options()
options_schema.update({
            'contract': 'optionroot',
            'date': 'quotedate'})

Out [9]:

Schema([Field(name='underlying', mapping='underlying'), Field(name='underlying_last', mapping='underlying_last'), Field(name='date', mapping='quotedate'), Field(name='contract', mapping='optionroot'), Field(name='type', mapping='type'), Field(name='expiration', mapping='expiration'), Field(name='strike', mapping='strike'), Field(name='bid', mapping='bid'), Field(name='ask', mapping='ask'), Field(name='volume', mapping='volume'), Field(name='open_interest', mapping='open_interest')])

In the last cell we updated the schema so that the contract and date columns are mapped to optionroot and quotedate, as they are in our data. This isn't needed in our case, since the default schema used by the backtester already uses this mapping; in general though, to use a custom schema pass it to Stock or HistoricalOptionsData when loading the csv (schema=None uses the default schema).

In [10]:

options_data = HistoricalOptionsData(OPTIONS_DATA, schema=None)
schema = options_data.schema

In [11]:

options_data.head()

Out [11]:

	underlying	underlying_last	exchange	optionroot	optionext	type	expiration	quotedate	strike	last	...	ask	openinterest	impliedvol	delta	theta	optionalias	dte
0	SPX	2257.83	*	SPX170120C00100000	NaN	call	2017-01-20	2017-01-03	100	2161.50	...	2158.7	5615	0.1214	1.0	-1.0239	SPX170120C00100000	17
1	SPX	2257.83	*	SPX170120C00200000	NaN	call	2017-01-20	2017-01-03	200	2060.95	...	2058.8	6355	0.1214	1.0	-2.0479	SPX170120C00200000	17
2	SPX	2257.83	*	SPX170120C00300000	NaN	call	2017-01-20	2017-01-03	300	1960.45	...	1958.8	2000	0.1214	1.0	-3.0718	SPX170120C00300000	17
3	SPX	2257.83	*	SPX170120C00400000	NaN	call	2017-01-20	2017-01-03	400	1760.55	...	1858.9	25	0.1214	1.0	-4.0957	SPX170120C00400000	17
4	SPX	2257.83	*	SPX170120C00500000	NaN	call	2017-01-20	2017-01-03	500	1749.33	...	1758.9	3813	0.1214	1.0	-5.1197	SPX170120C00500000	17

5 rows × 21 columns

For the options part of the backtest, we create a Strategy to determine when to enter on certain contracts. Remember to pass the options schema when creating it.

In [12]:

long_straddle = Strategy(schema)

Each strategy consists of one or more legs created with StrategyLeg. A leg consists of a type of contract (put or call) and a direction to enter (buy or sell). Our long straddle will have two legs, which will buy a call and a put respectively.

In [13]:

leg1 = StrategyLeg("leg_1", schema, option_type=Type.CALL, direction=Direction.BUY)

leg2 = StrategyLeg("leg_2", schema, option_type=Type.PUT, direction=Direction.BUY)

Legs have entry and exit filters, to determine the conditions a contract from that leg has to satisfy to select it for entry/exit. Filters are created with the following syntax, using the options schema to refer to column names (such as underlying, dte or strike).

In [14]:

leg1.entry_filter = (schema.underlying == "SPX") & (schema.dte >= 31) & (schema.dte <= 60) & (schema.strike >= schema.underlying_last * 0.95) & (schema.strike <= schema.underlying_last * 1.05)
leg2.entry_filter = (schema.underlying == "SPX") & (schema.dte >= 31) & (schema.dte <= 60) & (schema.strike >= schema.underlying_last * 0.95) & (schema.strike <= schema.underlying_last * 1.05)

leg1.exit_filter = (schema.dte <= 60)
leg2.exit_filter = (schema.dte <= 60)

long_straddle.add_legs([leg1, leg2]);

In this case, the first leg will look for contracts with SPX as the underlying which have between 31 and 60 days to expiration (dte) and are at the money (ATM), i.e, their strike price is close to their spot. After entering, the exit filter will exit contracts having 60 or less dte, which for this given strategy means we will always exit contracts. This effectively means we are buying ATM put and call contracts, keeping them for a month and then selling them.

To put it all together, create a Backtest object and pass it a map specifying the percentage allocated to stocks, options and cash. Then set the stock portfolio, the options strategy and the data and run it.

In [15]:

bt = Backtest({'stocks': 0.97, 'options': 0.03, 'cash': 0})
bt.stocks = stocks
bt.options_strategy = long_straddle
bt.options_data = options_data
bt.stocks_data = stock_data

The run method takes the initial capital (default value 1.000.000) and rebalance_freq as arguments. This last one defines how often, in months, a rebalancing of the portfolio is made, so that a value of 1 means a monthly rebalancing and 6 a bi-annual one. Its value defaults to 0, meaning no rebalancing done.

In [16]:

bt.run(rebalance_freq=1)

Out [16]:

0% [██████████████████████████████] 100% | ETA: 00:00:00
Total time elapsed: 00:00:04

	leg_1							leg_2							totals
	contract	underlying	expiration	type	strike	cost	order	contract	underlying	expiration	type	strike	cost	order	cost	qty	date
0	SPX170217C02150000	SPX	2017-02-17	call	2150	11800.0	Order.BTO	SPX170217P02150000	SPX	2017-02-17	put	2150	1220.0	Order.BTO	13020.0	2.0	2017-01-03
1	SPX170217C02150000	SPX	2017-02-17	call	2150	-12740.0	Order.STC	SPX170217P02150000	SPX	2017-02-17	put	2150	-160.0	Order.STC	-12900.0	2.0	2017-02-01
2	SPX170317C02170000	SPX	2017-03-17	call	2170	11660.0	Order.BTO	SPX170317P02170000	SPX	2017-03-17	put	2170	1010.0	Order.BTO	12670.0	2.0	2017-02-01
3	SPX170317C02170000	SPX	2017-03-17	call	2170	-22360.0	Order.STC	SPX170317P02170000	SPX	2017-03-17	put	2170	-55.0	Order.STC	-22415.0	2.0	2017-03-01
4	SPX170421C02280000	SPX	2017-04-21	call	2280	12610.0	Order.BTO	SPX170421P02280000	SPX	2017-04-21	put	2280	1110.0	Order.BTO	13720.0	2.0	2017-03-01
5	SPX170421C02280000	SPX	2017-04-21	call	2280	-8120.0	Order.STC	SPX170421P02280000	SPX	2017-04-21	put	2280	-285.0	Order.STC	-8405.0	2.0	2017-04-03
6	SPX170519C02245000	SPX	2017-05-19	call	2245	12420.0	Order.BTO	SPX170519P02245000	SPX	2017-05-19	put	2245	1040.0	Order.BTO	13460.0	2.0	2017-04-03
7	SPX170519C02245000	SPX	2017-05-19	call	2245	-14080.0	Order.STC	SPX170519P02245000	SPX	2017-05-19	put	2245	-120.0	Order.STC	-14200.0	2.0	2017-05-01
8	SPX170616C02270000	SPX	2017-06-16	call	2270	12470.0	Order.BTO	SPX170616P02270000	SPX	2017-06-16	put	2270	820.0	Order.BTO	13290.0	2.0	2017-05-01
9	SPX170616C02270000	SPX	2017-06-16	call	2270	-15690.0	Order.STC	SPX170616P02270000	SPX	2017-06-16	put	2270	-65.0	Order.STC	-15755.0	2.0	2017-06-01
10	SPX170721C02310000	SPX	2017-07-21	call	2310	12970.0	Order.BTO	SPX170721P02310000	SPX	2017-07-21	put	2310	860.0	Order.BTO	13830.0	2.0	2017-06-01
11	SPX170721C02310000	SPX	2017-07-21	call	2310	-11650.0	Order.STC	SPX170721P02310000	SPX	2017-07-21	put	2310	-220.0	Order.STC	-11870.0	2.0	2017-07-03
12	SPX170818C02310000	SPX	2017-08-18	call	2310	12590.0	Order.BTO	SPX170818P02310000	SPX	2017-08-18	put	2310	1060.0	Order.BTO	13650.0	2.0	2017-07-03
13	SPX170818C02310000	SPX	2017-08-18	call	2310	-16190.0	Order.STC	SPX170818P02310000	SPX	2017-08-18	put	2310	-120.0	Order.STC	-16310.0	2.0	2017-08-01
14	SPX170915C02355000	SPX	2017-09-15	call	2355	12700.0	Order.BTO	SPX170915P02355000	SPX	2017-09-15	put	2355	790.0	Order.BTO	13490.0	2.0	2017-08-01
15	SPX170915C02355000	SPX	2017-09-15	call	2355	-11830.0	Order.STC	SPX170915P02355000	SPX	2017-09-15	put	2355	-100.0	Order.STC	-11930.0	2.0	2017-09-01
16	SPX171020C02355000	SPX	2017-10-20	call	2355	13230.0	Order.BTO	SPX171020P02355000	SPX	2017-10-20	put	2355	1020.0	Order.BTO	14250.0	2.0	2017-09-01
17	SPX171020C02355000	SPX	2017-10-20	call	2355	-17230.0	Order.STC	SPX171020P02355000	SPX	2017-10-20	put	2355	-95.0	Order.STC	-17325.0	2.0	2017-10-02
18	SPX171117C02405000	SPX	2017-11-17	call	2405	13220.0	Order.BTO	SPX171117P02405000	SPX	2017-11-17	put	2405	780.0	Order.BTO	14000.0	2.0	2017-10-02
19	SPX171117C02405000	SPX	2017-11-17	call	2405	-16950.0	Order.STC	SPX171117P02405000	SPX	2017-11-17	put	2405	-110.0	Order.STC	-17060.0	2.0	2017-11-01
20	SPX171215C02455000	SPX	2017-12-15	call	2455	13090.0	Order.BTO	SPX171215P02455000	SPX	2017-12-15	put	2455	850.0	Order.BTO	13940.0	2.0	2017-11-01
21	SPX171215C02455000	SPX	2017-12-15	call	2455	-18550.0	Order.STC	SPX171215P02455000	SPX	2017-12-15	put	2455	-120.0	Order.STC	-18670.0	2.0	2017-12-01
22	SPX180119C02515000	SPX	2018-01-19	call	2515	14480.0	Order.BTO	SPX180119P02515000	SPX	2018-01-19	put	2515	1070.0	Order.BTO	15550.0	2.0	2017-12-01

When done, the backtester returns a trade log with all the option trades. There is also a balance sheet with daily returns, capital and other information for each stock symbol and options.

In [17]:

bt.balance

Out [17]:

	total capital	cash	VOO	TUR	RSX	EWY	EWS	VTIP	TLT	BWX	...	IAU	VNQI	options qty	calls capital	puts capital	stocks qty	options capital	stocks capital	% change	accumulated return
2017-01-02	1.000000e+06	1000000.000000	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	...	NaN	NaN	NaN	NaN	NaN	NaN	0.0	0.000000e+00	NaN	NaN
2017-01-03	9.991600e+05	4252.203989	96913.915483	48486.088959	48493.996041	48485.711480	48496.962760	96970.393641	193897.445209	96996.896604	...	145492.92	96981.885738	2.0	22960.0	2240.0	34536.0	25200.0	9.697078e+05	-0.000840	0.999160
2017-01-04	1.005354e+06	4252.203989	97485.817277	48563.271527	48808.018217	48817.743929	49267.903052	97009.844574	194642.955279	97299.303141	...	146144.77	98265.641145	2.0	24600.0	1620.0	34536.0	26220.0	9.748815e+05	0.006199	1.005354
2017-01-05	1.013846e+06	4252.203989	97410.813763	48362.596851	48606.146819	49302.331828	50207.486533	97128.197374	197689.822522	98395.526836	...	148361.06	99432.691515	2.0	24380.0	1600.0	34536.0	25980.0	9.836141e+05	0.008447	1.013846
2017-01-06	1.010046e+06	4252.203989	97790.519052	47884.064932	48269.694487	48763.900829	49894.292040	96970.393641	195874.667569	97715.112129	...	147318.10	99335.437317	2.0	26020.0	1240.0	34536.0	27260.0	9.785336e+05	-0.003749	1.010046
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
2017-12-21	1.172150e+06	4311.464660	114876.694450	62052.589006	57247.456248	55795.201266	56439.121909	112930.078174	223250.588763	111696.704128	...	168071.82	114300.129718	2.0	34780.0	500.0	36059.0	35280.0	1.132559e+06	0.002874	1.172150
2017-12-22	1.175638e+06	4311.464660	114853.445706	62409.383959	58044.851172	56275.993827	56351.551511	112884.385243	223555.184269	111934.694264	...	169037.75	115033.070138	2.0	34280.0	520.0	36059.0	34800.0	1.136527e+06	0.002976	1.175638
2017-12-26	1.178685e+06	4311.464660	114720.416392	62721.579544	58594.778705	56066.616421	56285.873712	113002.044540	224218.127430	111934.694264	...	170003.68	115206.661291	2.0	34160.0	470.0	36059.0	34630.0	1.139744e+06	0.002592	1.178685
2017-12-27	1.184637e+06	4311.464660	114785.826746	62587.781436	58292.318562	56904.126044	56504.799708	113048.432900	227138.660814	112410.674538	...	170831.62	116016.753334	2.0	34180.0	520.0	36059.0	34700.0	1.145626e+06	0.005050	1.184637
2017-12-28	1.189650e+06	4311.464660	115000.746480	64460.954942	58264.822185	57997.541385	56723.725705	113071.627080	226941.569604	112371.009515	...	171659.56	116768.981660	2.0	34140.0	420.0	36059.0	34560.0	1.150779e+06	0.004231	1.189650

251 rows × 21 columns

In [18]:

bt.balance['total capital'].plot();

In [19]:

bt.balance[[stock.symbol for stock in stocks]].plot();

Pass the balance dataframe to the functions returns_chart, returns_histogram and monthly_returns_heatmap for better visualization.

In [20]:

returns_chart(bt.balance)

Out [20]:

In [21]:

monthly_returns_heatmap(bt.balance)

Out [21]:

In [22]:

returns_histogram(bt.balance)

Out [22]:

The backtest also supports an sma option. If the argument sma_days is passed to the run method with a value of n, on each rebalancing day the backtest will only buy shares of a stock symbol if its current price is above the sma of the last n days. To illustrate, we run the same backtest as before but with sma_days = 30.

In [23]:

bt.run(rebalance_freq=1, sma_days=30);

0% [██████████████████████████████] 100% | ETA: 00:00:00
Total time elapsed: 00:00:05

In [26]:

bt.balance[[stock.symbol for stock in stocks]].plot();

Notice how the capital held in certain symbols sometimes drops to zero, meaning on those months we did not buy shares of those symbols because their price was below the sma.

Finally, the daily returns column (% change) can be passed to the Quantopian library pyfolio to obtain summary statistics and a few plots of the returns.

In [25]:

pf.create_returns_tear_sheet(returns =  bt.balance['% change'].dropna())

	Backtest
Start date	2017-01-03
End date	2017-12-28
Total months	11
Annual return	13.3%
Cumulative returns	13.2%
Annual volatility	4.9%
Sharpe ratio	2.59
Calmar ratio	5.99
Stability	0.92
Max drawdown	-2.2%
Omega ratio	1.53
Sortino ratio	4.13
Skew	-0.02
Kurtosis	0.67
Tail ratio	1.10
Daily value at risk	-0.6%

/Users/jrchatruc/anaconda3/lib/python3.7/site-packages/numpy/core/fromnumeric.py:61: FutureWarning: 
The current behaviour of 'Series.argmin' is deprecated, use 'idxmin'
instead.
The behavior of 'argmin' will be corrected to return the positional
minimum in the future. For now, use 'series.values.argmin' or
'np.argmin(np.array(values))' to get the position of the minimum
row.
  return bound(*args, **kwds)

Worst drawdown periods	Net drawdown in %	Peak date	Valley date	Recovery date	Duration
0	2.22	2017-06-06	2017-07-06	2017-08-28	60
1	1.82	2017-09-07	2017-09-27	2017-10-27	37
2	1.80	2017-03-01	2017-03-09	2017-04-24	39
3	1.10	2017-01-17	2017-01-19	2017-01-31	11
4	0.98	2017-05-15	2017-05-18	2017-05-22	6

1.2 MiB Raw Blame History Unescape Escape

1.2 MiB

Raw Blame History