Pass dates to strategy

Maxim Korobov

My idea now is to pass earning dates (plus day before than and two after) of particular company into strategy to avoid any trading in that high volatility dates.

Dates were taken from the public source. So I should pass list of dates into strategy. Note that there is only date column, no OLHC data. Could you please show generic DataBase subclass example, which could be passed into strategy and later used as data[x]?

Filter looks simpler, but currently the idea isn't proved, maybe it's better to reverse it and trade on earning days only :) . So earning dates could be used in strategy. Also note that length of earning days list is smaller that main data.

backtrader

You need to use linesoverride. See: Blog - Escape from OHLC Land

Maxim Korobov

What if data is in operating memory, not in csv file on disk?

Maxim Korobov

Something like this:

class EarningFeed(feed.DataBase):
	linesoverride = True
	lines = ('earning_date', 'datetime')

	def __init__(self, earning_dates):
		super().__init__()
		self.earning_dates = earning_dates
		self._idx = 0

	def start(self):
		super().start()

	def _load(self):
		self._idx += 1

		if self._idx >= len(self.earning_dates):
			return False

		self.lines.earning_date[0] = date2num(self.earning_dates[0])
		self.lines.datetime[0] = date2num(self.earning_dates[0])

		return True

One issue I stucked on is cerebro run - it stops too early. Seem like it want all the datetime rows like on the main datafeed.

backtrader

Without knowing what too early actually implies/means in that context, the problem with your data feed is that it only delivers something at fixed points in time.

And without having put it to the test, the first time your strategy's next is called will be the 1st time that the earnings data feed produces a valued. Because at that precise moment in time, the data feed meets its minimum period requirement (actually 1 for data feeds)

Maxim Korobov

Roger.

Fixed by adding fake datetimes from the start data date parameters, which is first date in main data feed:

class EarningFeed(feed.DataBase):
    linesoverride = True
    lines = ('earning_date', 'datetime')

    def __init__(self, earning_dates, start_data_date):
        super().__init__()
        self.start_data_date = start_data_date
        self.earning_dates = earning_dates
        self.idx = 0

    def start(self):
        super().start()

    def _load(self):
        self.idx += 1

        if self.idx >= len(self.earning_dates):
            return False

        date = self.start_data_date + dt.timedelta(days=self.idx)
        self.lines.datetime[0] = date2num(date)
        self.lines.earning_date[0] = date2num(self.earning_dates[self.idx])

        return True

Next issue is plotting data. Seems like one should provide Y data:

Traceback (most recent call last):
...
  File "C:\Users\Home\Anaconda3\lib\site-packages\backtrader\cerebro.py", line 659, in plot
	start=start, end=end)
  File "C:\Users\Home\Anaconda3\lib\site-packages\backtrader\plot\plot.py", line 205, in plot
	self.plotdata(data, self.dplotsover[data])
  File "C:\Users\Home\Anaconda3\lib\site-packages\backtrader\plot\plot.py", line 625, in plotdata
	color=color, label=datalabel)
  File "C:\Users\Home\Anaconda3\lib\site-packages\backtrader\plot\finance.py", line 594, in plot_lineonclose
	**kwargs)
  File "C:\Users\Home\Anaconda3\lib\site-packages\backtrader\plot\finance.py", line 534, in __init__
	ax.add_line(self.loc)
  File "C:\Users\Home\Anaconda3\lib\site-packages\matplotlib\axes\_base.py", line 1787, in add_line
	self._update_line_limits(line)
  File "C:\Users\Home\Anaconda3\lib\site-packages\matplotlib\axes\_base.py", line 1809, in _update_line_limits
	path = line.get_path()
  File "C:\Users\Home\Anaconda3\lib\site-packages\matplotlib\lines.py", line 989, in get_path
	self.recache()
  File "C:\Users\Home\Anaconda3\lib\site-packages\matplotlib\lines.py", line 696, in recache
	raise RuntimeError('xdata and ydata must be the same length')
RuntimeError: xdata and ydata must be the same length

Hope to solve this tomorrow.

backtrader

Try running cerebro with runonce=False.

In any case and without knowing the use case in detail and seeing that the earning_dates seem to already be in a preprocessed array.

Wouldn't it be better to have those earnings as an indicator applied to the data feed?

With that approach you'd only need to do something like this in next:

if self.data.datetime[0] == self.earning_dates[self._idx]:  # here earning_dates already contains date2num'd timestamps
    do_something()
    self.lines.mylinename[0] = 1.0  # for example, any other value
    self._idx += 1  # update index

And that would automatically plot synchronized with the target data feed.

Maxim Korobov

Thanks for advice and overall help. It was easy to build and use indicator.

Final result (contains references to external code - main project):

import backtrader as bt
from backtrader import num2date

from main.lab.data.earning_dates import EarningDates


class EarningIndicator(bt.Indicator):

	params = (
		("earning", None),
		("skip_before", 1),
		("skip_after", 2),
	)

	lines = ("earning_period",)

	def __init__(self, *args, **kwargs):
		super().__init__(*args, **kwargs)

		self.skip_days = []

		a = EarningDates(0, self.params.earning, self.params.skip_before, self.params.skip_after)

		self.skip_days = a.get_skip_dates()

	def next(self):
		current_date = num2date(self.data.datetime[0]).date()
		if current_date in self.skip_days:
			self.lines.earning_period[0] = 1.0
		else:
			self.lines.earning_period[0] = 0