ematvey / pybacktest Goto Github PK

Hi! First of all great library. I was looking for something like this as I am looking for a backtester for http://docs.kloudtrader.com. It's always better to extend something instead of reinventing the wheel but I don't see any License. So i wanted to seek permission to extend this library! Lemme know if this could be possible.
Thanks!

Hi there,

To properly adjust your historical OHLC data, you need to use a proportional adjustment. Here's the function that does just that:

def adjust_ohlc(data):
    """
    Adjust OHLC prices for splits and dividends by using 
    the Adjusted column in Yahoo!.
    """
    ratio = data['AC']/data['C']
    data['UC'] = data['C']  # add an unadjusted close column
    data[['O', 'H', 'L', 'C']] *= ratio
    return data.drop(['AC'], axis=1)

Note that it does not adjust the Volume.

Any suggestions for implementing stop loss and profit target exits? I would be willing to write and contribute any changes to the code with some input.

Hi, I got a little confused about the difference between positions and pos in trades. I see that there are two shift in pos in trades. Can you tell me why? Thanks very much.

When running this:

import pybacktest
ohlc = pybacktest.load_from_yahoo('SPY')

Raises error:

---------------------------------------------------------------------------
OverflowError                             Traceback (most recent call last)
<ipython-input-14-09359e4057e9> in <module>()
----> 1 ohlc = pybacktest.load_from_yahoo('SPY')

~/.pyenv/versions/3.5.3/envs/acdt/lib/python3.5/site-packages/pybacktest/data.py in load_from_yahoo(ticker, start, adjust_close)
     24              for t in ticker})
     25 
---> 26     data = dr.DataReader(ticker, data_source='yahoo', start=start)
     27     r = data['Adj Close'] / data['Close']
     28     ohlc_cols = ['Open', 'High', 'Low', 'Close']

~/.pyenv/versions/3.5.3/envs/acdt/lib/python3.5/site-packages/pandas_datareader/data.py in DataReader(name, data_source, start, end, retry_count, pause, session, access_key)
    119                                 adjust_price=False, chunksize=25,
    120                                 retry_count=retry_count, pause=pause,
--> 121                                 session=session).read()
    122 
    123     elif data_source == "yahoo-actions":

~/.pyenv/versions/3.5.3/envs/acdt/lib/python3.5/site-packages/pandas_datareader/yahoo/daily.py in read(self)
    113         """ read one data from specified URL """
    114         try:
--> 115             df = super(YahooDailyReader, self).read()
    116             if self.ret_index:
    117                 df['Ret_Index'] = _calc_return_index(df['Adj Close'])

~/.pyenv/versions/3.5.3/envs/acdt/lib/python3.5/site-packages/pandas_datareader/base.py in read(self)
    179         if isinstance(self.symbols, (compat.string_types, int)):
    180             df = self._read_one_data(self.url,
--> 181                                      params=self._get_params(self.symbols))
    182         # Or multiple symbols, (e.g., ['GOOG', 'AAPL', 'MSFT'])
    183         elif isinstance(self.symbols, DataFrame):

~/.pyenv/versions/3.5.3/envs/acdt/lib/python3.5/site-packages/pandas_datareader/yahoo/daily.py in _get_params(self, symbol)
     97 
     98     def _get_params(self, symbol):
---> 99         unix_start = int(time.mktime(self.start.timetuple()))
    100         day_end = self.end.replace(hour=23, minute=59, second=59)
    101         unix_end = int(time.mktime(day_end.timetuple()))

OverflowError: mktime argument out of range

Hey, is it possible to change default position size to percent of equity or set amount of dollars?

Shouldn't the sharpe ratio calc be

return (d.mean() / d.std()) * (252**0.5)

instead of

return (d.mean() / d.std()) ** (252**0.5)

This is an issue with the pictured hello world code (bottom left).

Actually, here is the shortest MWE:

import matplotlib
import matplotlib.pyplot as plt
import pybacktest
import pandas as pd

short_ma = 50
long_ma = 200

ohlc = pybacktest.load_from_yahoo('AAPL', start=2000)
ohlc.tail()

ms = ohlc.C.rolling(short_ma).mean()
ml = ohlc.C.rolling(long_ma).mean()

buy = cover = (ms > ml) & (ms.shift() < ml.shift())  # ma cross up
sell = short = (ms < ml) & (ms.shift() > ml.shift())  # ma cross down

bt = pybacktest.Backtest(locals(), 'ma_cross')

print(bt.summary())

matplotlib.rcParams['figure.figsize'] = (15.0, 8.0)

bt.plot_equity()

Приветствую. Подскажите как я могу посчитать эквити в процентах?
Как мне кажется лучше бы код что отвечает за эквити переделать в проценты, а то в статистике данные неверные вылезают. Или я ошибаюсь?
Спасибо

For the pythonically challenged... could someone please modify the example notebook to use 4 different buy, cover, short,and sell rules and to show the backtest fuction the long way..

pybacktest.Backtest(self, dataobj, name='Unknown', signal_fields=('buy', 'sell', 'short', 'cover'), price_fields=('buyprice', 'sellprice', 'shortprice', 'coverprice'))

with 'dataobj', name, signal_fields', and 'price_filelds'. the 'locals()' function is making it hard for me to figure this out.. thanks in advance.

you should add the pyquant package to your setup file since it causes an error.

The tutorial mentions figsize.
What module does it belong to?

When testing with hourly data instead of the daily on the example app (ms/ml) and calling the summary, I get the following error:

Traceback (most recent call last):
  File "pyback1.py", line 75, in <module>
    print(bt.summary())
  File "/usr/local/lib/python3.7/site-packages/pybacktest/backtest.py", line 174, in summary
    print((yaml.dump(self.report, allow_unicode=True, default_flow_style=False)))
  File "/usr/local/lib/python3.7/site-packages/cached_property.py", line 35, in __get__
    value = obj.__dict__[self.func.__name__] = self.func(obj)
  File "/usr/local/lib/python3.7/site-packages/pybacktest/backtest.py", line 164, in report
    return pybacktest.performance.performance_summary(self.equity)
  File "/usr/local/lib/python3.7/site-packages/cached_property.py", line 35, in __get__
    value = obj.__dict__[self.func.__name__] = self.func(obj)
  File "/usr/local/lib/python3.7/site-packages/pybacktest/backtest.py", line 152, in equity
    return pybacktest.parts.trades_to_equity(self.trades)
  File "/usr/local/lib/python3.7/site-packages/cached_property.py", line 35, in __get__
    value = obj.__dict__[self.func.__name__] = self.func(obj)
  File "/usr/local/lib/python3.7/site-packages/pybacktest/backtest.py", line 142, in trades
    assert p.index.tz == tp.index.tz, "Cant operate on singals and prices " \
AttributeError: 'Index' object has no attribute 'tz'

The data looks similar, with the date format being more detailed.

I was using pd.to_datetime on a csv (without a timezone information). If there is no TimeZone information it would be nice to have a better error message.

Solution: Add a Timezone, pd.to_datetime(x, utc=True) in my case

Hey thank you for this simple backtester I want to do a backtest for a multiple indicator strategy

Eg:
Buy: Short_ SMA > Long_SMA & RSI < 20 & Short_EMA > Long_EMA
Sell: Short_ SMA < Long_SMA & RSI > 80 & Short_EMA < Long_EMA

I have gone through the example code and I have a confusion in here bt = pybacktest.Backtest(locals(), 'ma_cross'). Instead of using locals() I want to use my own custom dict. What format as well as values should I use in this dict?

Thank you for your help

Hi there,

would it be possible to test only buy signals to evaluate buy signals strength/ performances?

Hello!
When i wanted install a package
"pip install git+https://github.com/ematvey/pybacktest.git"
I got a error:
Fatal: unable to access 'https://github.com/ematvey/pybacktest.git/': OpenSSL SSL_connect: SSL_ERROR_SYSCALL in connection to github.com:443

What am i doing wrong?

I like your whole approach on performing efficients backtest via pandas. But could you provide more documentation regarding your performance statistics and reporting file?

Hello, I noticed Optimizer class, is it working, or is it unfinished?

I have gone through the example code and I have a confusion in here bt = pybacktest.Backtest(locals(), 'ma_cross'). Instead of using locals() I want to use my own custom dict. What format as well as values should I use in this dict?

Thank you for your help

I used it before and it worked. now, I am using python 2.7.3, pandas 0.13.1. it gives no result by calling summary(). positions, prices, report, equity, trades are empty. but trade_price is fine.

pls, correct code
+
here and in nearby
pf = PF = lambda eqd: abs(eqd[eqd > 0].sum() / eqd[eqd < 0].sum())
get error when eqd[eqd < 0].sum() == 0

I am using Python 3.7 and Pandas 0.24

I get this error from holding_periods function.

def holding_periods(eqd):
    # rather crude, but will do
    return pd.Series(eqd.index.to_datetime(), index=eqd.index, dtype=object).diff().dropna()

Should it be?

def holding_periods(eqd):
    # rather crude, but will do
    return pd.Series(pd.to_datetime (eqd.index), index=eqd.index, dtype=object).diff().dropna()

Or it might be deprecated in latest versions of Pandas:

https://pandas.pydata.org/pandas-docs/version/0.21/generated/pandas.Timestamp.to_datetime.html

Timestamp.to_datetime()
DEPRECATED: use to_pydatetime() instead.

Convert a Timestamp object to a native Python datetime object.

AttributeError Traceback (most recent call last)
in
----> 1 bt.summary()

C:\ProgramData\Anaconda3\lib\site-packages\pybacktest\backtest.py in summary(self)
172 print(s)
173 print(('-' * len(s) + '\n'))
--> 174 print((yaml.dump(self.report, allow_unicode=True, default_flow_style=False)))
175 print(('-' * len(s)))
176

C:\ProgramData\Anaconda3\lib\site-packages\cached_property.py in get(self, obj, cls)
33 return self._wrap_in_coroutine(obj)
34
---> 35 value = obj.dict[self.func.name] = self.func(obj)
36 return value
37

C:\ProgramData\Anaconda3\lib\site-packages\pybacktest\backtest.py in report(self)
162 @cached_property
163 def report(self):
--> 164 return pybacktest.performance.performance_summary(self.equity)
165
166 def summary(self):

C:\ProgramData\Anaconda3\lib\site-packages\pybacktest\performance.py in performance_summary(equity_diffs, quantile, precision)
89 if len(eqd) == 0:
90 return {}
---> 91 hold = holding_periods(equity_diffs)
92
93 return _format_out({

C:\ProgramData\Anaconda3\lib\site-packages\pybacktest\performance.py in holding_periods(eqd)
68 def holding_periods(eqd):
69 # rather crude, but will do
---> 70 return pd.Series(eqd.index.to_datetime(), index=eqd.index, dtype=object).diff().dropna()
71
72

AttributeError: 'DatetimeIndex' object has no attribute 'to_datetime'

Thank you for open-sourcing this nice package. I was wondering if there is a plan to release it to pypi, so it would be more accessible. Downloading/installing would take just one command.