Skip to content

Gold vs SP500

Sometimes getting hints about where backtrader is in use helps understanding what people may be looking for and using the platform for.

The reference:

It is a post (Spanish) analyzing two ETFs: GLD vs SPY (effectively Gold vs S&P500)

Without going into the translation, let’s concentrate on the important points for backtrader:

  • Adding a Correlation indicator. For the sake of it PearsonR was chosen.

    And for the feat of creating it and instead of coding it from scratch, a sample of how to do it from a scipy function is done. The code

    class PearsonR(bt.ind.PeriodN):
    _mindatas = 2  # hint to the platform

    lines = ('correlation',)
    params = (('period', 20),)

    def next(self):
        c, p, = scipy.stats.pearsonr(self.data0.get(size=self.p.period),
                                     self.data1.get(size=self.p.period))

        self.lines.correlation[0] = c

  • Adding rolling logarithmic returns

    The platform already had an analyzer with logarithmic returns, but not rolling.

    The analyzer LogReturnsRolling has been added, which takes a timeframe parameter (and compression) to use a different timeframe than that of the data (if needed be)

    And together with it and for visualization (using the analyzer internally) a LogReturns observer

  • Allowing data on data plotting (easily). Just like this

        # Data feeds
    data0 = YahooData(dataname=args.data0, **kwargs)
    # cerebro.adddata(data0)
    cerebro.resampledata(data0, timeframe=bt.TimeFrame.Weeks)

    data1 = YahooData(dataname=args.data1, **kwargs)
    # cerebro.adddata(data1)
    cerebro.resampledata(data1, timeframe=bt.TimeFrame.Weeks)
    data1.plotinfo.plotmaster = data0

    Just using plotmaster=data0 will plot data1 on data0

Plotting the moving averages on its own axis and on each other was supported from the very start of the platform.

The analyzer and observer have also been added to the platform, together with a sample with the same defaults as those from the blog post.

Running the sample:

$ ./gold-vs-sp500.py --cerebro stdstats=False --plot volume=False

Note

stdstats=False and volume=False are there to reduce clutter in the chart by removing some of the usual things like CashValue observer and the volume subplots.

Produces a chart which mimics most of the output from the article.

image

Not included:

  • The charts creating distributions of returns.

    They wouldn’t fit in the chart which has a datetime based x axis.

But having those distributions may come.

Sample Usage

$ ./gold-vs-sp500.py --help
usage: gold-vs-sp500.py [-h] [--data0 TICKER] [--data1 TICKER] [--offline]
                        [--fromdate FROMDATE] [--todate TODATE]
                        [--cerebro kwargs] [--broker kwargs] [--sizer kwargs]
                        [--strat kwargs] [--plot [kwargs]] [--myobserver]

Gold vs SP500 from https://estrategiastrading.com/oro-bolsa-estadistica-con-
python/

optional arguments:
  -h, --help           show this help message and exit
  --data0 TICKER       Yahoo ticker to download (default: SPY)
  --data1 TICKER       Yahoo ticker to download (default: GLD)
  --offline            Use the offline files (default: False)
  --fromdate FROMDATE  Date[time] in YYYY-MM-DD[THH:MM:SS] format (default:
                       2005-01-01)
  --todate TODATE      Date[time] in YYYY-MM-DD[THH:MM:SS] format (default:
                       2016-01-01)
  --cerebro kwargs     kwargs in key=value format (default: )
  --broker kwargs      kwargs in key=value format (default: )
  --sizer kwargs       kwargs in key=value format (default: )
  --strat kwargs       kwargs in key=value format (default: )
  --plot [kwargs]      kwargs in key=value format (default: )

Sample Code

from __future__ import (absolute_import, division, print_function,
                        unicode_literals)

# Reference
# https://estrategiastrading.com/oro-bolsa-estadistica-con-python/

import argparse
import datetime

import scipy.stats

import backtrader as bt


class PearsonR(bt.ind.PeriodN):
    _mindatas = 2  # hint to the platform

    lines = ('correlation',)
    params = (('period', 20),)

    def next(self):
        c, p, = scipy.stats.pearsonr(self.data0.get(size=self.p.period),
                                     self.data1.get(size=self.p.period))

        self.lines.correlation[0] = c


class MACrossOver(bt.Strategy):
    params = (
        ('ma', bt.ind.MovAv.SMA),
        ('pd1', 20),
        ('pd2', 20),
    )

    def __init__(self):
        ma1 = self.p.ma(self.data0, period=self.p.pd1, subplot=True)
        self.p.ma(self.data1, period=self.p.pd2, plotmaster=ma1)
        PearsonR(self.data0, self.data1)


def runstrat(args=None):
    args = parse_args(args)

    cerebro = bt.Cerebro()

    # Data feed kwargs
    kwargs = dict()

    # Parse from/to-date
    dtfmt, tmfmt = '%Y-%m-%d', 'T%H:%M:%S'
    for a, d in ((getattr(args, x), x) for x in ['fromdate', 'todate']):
        if a:
            strpfmt = dtfmt + tmfmt * ('T' in a)
            kwargs[d] = datetime.datetime.strptime(a, strpfmt)

    if not args.offline:
        YahooData = bt.feeds.YahooFinanceData
    else:
        YahooData = bt.feeds.YahooFinanceCSVData

    # Data feeds
    data0 = YahooData(dataname=args.data0, **kwargs)
    # cerebro.adddata(data0)
    cerebro.resampledata(data0, timeframe=bt.TimeFrame.Weeks)

    data1 = YahooData(dataname=args.data1, **kwargs)
    # cerebro.adddata(data1)
    cerebro.resampledata(data1, timeframe=bt.TimeFrame.Weeks)
    data1.plotinfo.plotmaster = data0

    # Broker
    kwargs = eval('dict(' + args.broker + ')')
    cerebro.broker = bt.brokers.BackBroker(**kwargs)

    # Sizer
    kwargs = eval('dict(' + args.sizer + ')')
    cerebro.addsizer(bt.sizers.FixedSize, **kwargs)

    # Strategy
    if True:
        kwargs = eval('dict(' + args.strat + ')')
        cerebro.addstrategy(MACrossOver, **kwargs)

    cerebro.addobserver(bt.observers.LogReturns2,
                        timeframe=bt.TimeFrame.Weeks,
                        compression=20)

    # Execute
    cerebro.run(**(eval('dict(' + args.cerebro + ')')))

    if args.plot:  # Plot if requested to
        cerebro.plot(**(eval('dict(' + args.plot + ')')))


def parse_args(pargs=None):

    parser = argparse.ArgumentParser(
        formatter_class=argparse.ArgumentDefaultsHelpFormatter,
        description=(
            'Gold vs SP500 from '
            'https://estrategiastrading.com/oro-bolsa-estadistica-con-python/')
    )

    parser.add_argument('--data0', required=False, default='SPY',
                        metavar='TICKER', help='Yahoo ticker to download')

    parser.add_argument('--data1', required=False, default='GLD',
                        metavar='TICKER', help='Yahoo ticker to download')

    parser.add_argument('--offline', required=False, action='store_true',
                        help='Use the offline files')

    # Defaults for dates
    parser.add_argument('--fromdate', required=False, default='2005-01-01',
                        help='Date[time] in YYYY-MM-DD[THH:MM:SS] format')

    parser.add_argument('--todate', required=False, default='2016-01-01',
                        help='Date[time] in YYYY-MM-DD[THH:MM:SS] format')

    parser.add_argument('--cerebro', required=False, default='',
                        metavar='kwargs', help='kwargs in key=value format')

    parser.add_argument('--broker', required=False, default='',
                        metavar='kwargs', help='kwargs in key=value format')

    parser.add_argument('--sizer', required=False, default='',
                        metavar='kwargs', help='kwargs in key=value format')

    parser.add_argument('--strat', required=False, default='',
                        metavar='kwargs', help='kwargs in key=value format')

    parser.add_argument('--plot', required=False, default='',
                        nargs='?', const='{}',
                        metavar='kwargs', help='kwargs in key=value format')

    return parser.parse_args(pargs)


if __name__ == '__main__':
    runstrat()