Resample Tick Data - Backtrader
backtrader could already do resampling up from minute data. Accepting tick data was not a problem, by simply setting the 4 usual fields (open, high, low, close) to the tick value.
But passing the tick data to be resampled produced the same data again. As or release 1.1.11.88 this is no longer so. Now
-
TimeFrame (backtrader.TimeFrame) has been extended to contain constants and names for “Ticks”, “MicroSeconds” and “Seconds”
-
Resampling can manage the 3 aforementioned timeframes and sample them up.
Note
Because tick data is the lowest possible timeframe it can actually be “compressed” (n bars to 1 bar) but not be sampled up from a smallest timeframe.
The new release contains a small tickdata.csv
sample added to the sources data a new sample script resample-tickdata.py
to play with it.
Note
Updated the script to use the new Cerebro.resampledata
method which avoids the need to manually instantiate a backtrader.DataResampler
The default execution doesn’t touch the data:
$ ./resample-tickdata.py
Producing this chart:
Compressing 3 ticks to 1:
$ ./resample-tickdata.py --timeframe ticks --compression 3
Producing this chart:
After the compression we no longer have single “ticks” but “bars”.
Now compressing to seconds and 5 bars compression:
$ ./resample-tickdata.py --timeframe seconds --compression 5
With a new chart:
And finally to minutes. The sample data contains tick data from 4 different minutes (the last tick in the file is the only tick for the 4th minute):
$ ./resample-tickdata.py --timeframe minutes
With a 4 bars (at the top it can be seen the final price was 3069). The 4th bar is a single point given for this minute a single tick is present in the file.
The script usage:
$ ./resample-tickdata.py --help usage: resample-tickdata.py [-h] [--dataname DATANAME] [--timeframe {ticks,microseconds,seconds,minutes,daily,weekly,monthly}] [--compression COMPRESSION] Resampling script down to tick data optional arguments: -h, --help show this help message and exit --dataname DATANAME File Data to Load --timeframe {ticks,microseconds,seconds,minutes,daily,weekly,monthly} Timeframe to resample to --compression COMPRESSION Compress n bars into 1
And the code.
from __future__ import (absolute_import, division, print_function, unicode_literals) import argparse import backtrader as bt import backtrader.feeds as btfeeds def runstrat(): args = parse_args() # Create a cerebro entity cerebro = bt.Cerebro(stdstats=False) # Add a strategy cerebro.addstrategy(bt.Strategy) # Load the Data datapath = args.dataname or '../../datas/ticksample.csv' data = btfeeds.GenericCSVData( dataname=datapath, dtformat='%Y-%m-%dT%H:%M:%S.%f', timeframe=bt.TimeFrame.Ticks, ) # Handy dictionary for the argument timeframe conversion tframes = dict( ticks=bt.TimeFrame.Ticks, microseconds=bt.TimeFrame.MicroSeconds, seconds=bt.TimeFrame.Seconds, minutes=bt.TimeFrame.Minutes, daily=bt.TimeFrame.Days, weekly=bt.TimeFrame.Weeks, monthly=bt.TimeFrame.Months) # Resample the data data = cerebro.resampledata(data, timeframe=tframes[args.timeframe], compression=args.compression) # add a writer cerebro.addwriter(bt.WriterFile, csv=True) # Run over everything cerebro.run() # Plot the result cerebro.plot(style='bar') def parse_args(): parser = argparse.ArgumentParser( description='Resampling script down to tick data') parser.add_argument('--dataname', default='', required=False, help='File Data to Load') parser.add_argument('--timeframe', default='ticks', required=False, choices=['ticks', 'microseconds', 'seconds', 'minutes', 'daily', 'weekly', 'monthly'], help='Timeframe to resample to') parser.add_argument('--compression', default=1, required=False, type=int, help=('Compress n bars into 1')) return parser.parse_args() if __name__ == '__main__': runstrat()