@Nguyễn-Tài-Nguyên I have the same issue. Its a quedtion for @Ed-Bartosh that I hope can spare some time soon to help us out.

Have anyone done an indicator on this that they would be willing to share? Im interested in both simple and exponential smoothing.

Alternatively, please share some tips on how to approach this given that I have not gotten my head around to indicator development yet and are not experienced in sub-classing/inheritance in Python (with some tips I could likely hack my way through it)

in short any help would be highly appreciated. Thx in advance & Merry Christmas !

@søren-pallesen said in Indicator based on other indicators with 2 data feeds:

@backtrader thanks for the response. Just to be clear there are no problems plotting with multiple data feeds with different timeframes. It breaks when I add calculated indicators on the 2nd feed like:

        differ = self.ema_day - self.kama_day
        self.roc = btind.ROC(differ, period=6)

Adding the same calculated indicator on the first feed works and plots fine.

Getting an error when base an indicator on a calculation of two other indicators from a second data feed:

import backtrader as bt
import backtrader.indicators as btind

class myStrategy(bt.SignalStrategy):        
    def __init__(self):
        self.ema_day = btind.EMA(self.data1, period=15)
        self.kama_day = btind.KAMA(self.data1, period=25, fast=3, slow=30)
        differ = self.ema_day - self.kama_day
        self.roc = btind.ROC(differ, period=6)

    def next(self):
        pass

data = bt.feeds.GenericCSVData(
    dataname='./data/VXX-15m-2017-10-01us.csv',
    dtformat='%Y-%m-%d %H:%M:%S',
    openinterest=-1, 
    timeframe=bt.TimeFrame.Minutes, 
    compression=15
)    
    
data_day = bt.feeds.GenericCSVData(
    dataname='./data/VXX-1d-2017-10-01us.csv',
    dtformat='%Y-%m-%d',
    openinterest=-1, 
    timeframe=bt.TimeFrame.Days, 
    compression=1
)

cerebro = bt.Cerebro()
cerebro.addstrategy(myStrategy)
cerebro.adddata(data)
cerebro.adddata(data_day)

cerebro.run()
cerebro.plot()

And get this error: ValueError: x and y must have same first dimension, but have shapes (7542,) and (300,)

Here is the full message:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-1-6c0f97016110> in <module>
     34 
     35 cerebro.run()
---> 36 cerebro.plot()

~\Anaconda3\envs\IBbt\lib\site-packages\backtrader\cerebro.py in plot(self, plotter, numfigs, iplot, start, end, width, height, dpi, tight, use, **kwargs)
    989                 rfig = plotter.plot(strat, figid=si * 100,
    990                                     numfigs=numfigs, iplot=iplot,
--> 991                                     start=start, end=end, use=use)
    992                 # pfillers=pfillers2)
    993 

~\Anaconda3\envs\IBbt\lib\site-packages\backtrader\plot\plot.py in plot(self, strategy, figid, numfigs, iplot, start, end, **kwargs)
    226                         subinds=self.dplotsover[ind],
    227                         upinds=self.dplotsup[ind],
--> 228                         downinds=self.dplotsdown[ind])
    229 
    230             cursor = MultiCursor(

~\Anaconda3\envs\IBbt\lib\site-packages\backtrader\plot\plot.py in plotind(self, iref, ind, subinds, upinds, downinds, masterax)
    450                 xdata = np.array(xdata)[lplotmask]
    451 
--> 452             plottedline = pltmethod(xdata, lplotarray, **plotkwargs)
    453             try:
    454                 plottedline = plottedline[0]

~\Anaconda3\envs\IBbt\lib\site-packages\matplotlib\__init__.py in inner(ax, *args, **kwargs)
   1715                     warnings.warn(msg % (label_namer, func.__name__),
   1716                                   RuntimeWarning, stacklevel=2)
-> 1717             return func(ax, *args, **kwargs)
   1718         pre_doc = inner.__doc__
   1719         if pre_doc is None:

~\Anaconda3\envs\IBbt\lib\site-packages\matplotlib\axes\_axes.py in plot(self, *args, **kwargs)
   1370         kwargs = cbook.normalize_kwargs(kwargs, _alias_map)
   1371 
-> 1372         for line in self._get_lines(*args, **kwargs):
   1373             self.add_line(line)
   1374             lines.append(line)

~\Anaconda3\envs\IBbt\lib\site-packages\matplotlib\axes\_base.py in _grab_next_args(self, *args, **kwargs)
    402                 this += args[0],
    403                 args = args[1:]
--> 404             for seg in self._plot_args(this, kwargs):
    405                 yield seg
    406 

~\Anaconda3\envs\IBbt\lib\site-packages\matplotlib\axes\_base.py in _plot_args(self, tup, kwargs)
    382             x, y = index_of(tup[-1])
    383 
--> 384         x, y = self._xy_from_xy(x, y)
    385 
    386         if self.command == 'plot':

~\Anaconda3\envs\IBbt\lib\site-packages\matplotlib\axes\_base.py in _xy_from_xy(self, x, y)
    241         if x.shape[0] != y.shape[0]:
    242             raise ValueError("x and y must have same first dimension, but "
--> 243                              "have shapes {} and {}".format(x.shape, y.shape))
    244         if x.ndim > 2 or y.ndim > 2:
    245             raise ValueError("x and y can be no greater than 2-D, but have "

ValueError: x and y must have same first dimension, but have shapes (7542,) and (300,)

Its a resent install of BT (v1.9.67.122). I'm properly missing something simple ...Any ideas?

Reading the documentation/examples and studying community posts on resampling I still have a few questions. Mainly related to how compression in the feed relates to compression in resampledata and the specific role of sessionstart/sessionend.

Question 1: Im loading 1 hour bar data. Can I resample from 1 hour bars to 90 minute bars, like this?:

data = bt.feeds.GenericCSVData(
    dataname='./data/VXX-1h-2016-08-09us.csv', #1 hour bars
    timeframe=bt.TimeFrame.Minutes, 
    compression=60,
    sessionstart=datetime.time(9, 30),
    sessionend=datetime.time(15, 0))

#souce resolution is a higher timeframe than destination resolution - does that work?
cerebro.resampledata(data, timeframe=bt.TimeFrame.Minutes, compression=90)

Question 2: Im loading 15min bar data. Can i resample from 15min bars to 120min bars like this?:

data = bt.feeds.GenericCSVData(
    dataname='./data/VXX-15m-2016-08-09us.csv', #15 minute bars
    timeframe=bt.TimeFrame.Minutes, 
    compression=15,
    sessionstart=datetime.time(9, 30),
    sessionend=datetime.time(16, 0))

#souce resolution is divisible with destination resolution - is that a requirement?
cerebro.resampledata(data, timeframe=bt.TimeFrame.Minutes, compression=120)

Question 3: Im loading 15min bar data. Can i resample from 15min bars to 50min bars like this?:

data = bt.feeds.GenericCSVData(
    dataname='./data/VXX-15m-2016-08-09us.csv', #15 minute bars
    timeframe=bt.TimeFrame.Minutes, 
    compression=15,
    sessionstart=datetime.time(9, 30),
    sessionend=datetime.time(16, 0))

#souce resolution is indivisible with destination resolution - does that work?
cerebro.resampledata(data, timeframe=bt.TimeFrame.Minutes, compression=50)

Question 4: Using NASDAQ as an example trading session starts 9.30 and ends 16.00 and my feed data consists of 1 hour bars with first daily bar at 10.00 last daily bar at 15.00. Should sessionstart and sessionend be set:

    #the data's trading hours
    sessionstart=datetime.time(10, 0),
    sessionend=datetime.time(15, 0)

Or:

    #the exchange's actual trading hours
    sessionstart=datetime.time(9, 30),
    sessionend=datetime.time(16, 0)

Also if possible please explain the role og sessionstart and sessionend, like what does it do? And should always set them when loading data?

Thx in advance :-)

Thx a bunch for the pointers. Sounds very interesting poling Bitstamp, as im also trading there. Can you share the code you use to get data and construct candles on Bitstamp?

Anyone have tips for where to get good OHLCV candle data for backtesting on crypto currencies?

Are there any commercial offerings on the market?

Anyone have tips for where to get good OHLCV candle data for backtesting on crypto currencies?

Are there any commercial offerings on the market?

This worked fo me (it seems - still have more testing to do) in case anyone is interested ... :-)

                mainside = self.buy(transmit=False)
                lowside  = self.sell(price=self.data.close[0]*(1-self.params.sPerc), size=mainside.size, exectype=bt.Order.Stop, transmit=False, parent=mainside)
                highside = self.sell(price=self.data.close[0]*self.params.lPerc, size=mainside.size, exectype=bt.Order.Limit, transmit=False, parent=mainside)
                trailside = self.sell(size=mainside.size, exectype=bt.Order.StopTrail, trailpercent=self.params.tPerc, transmit=True, parent=mainside)

Incendently anyone know the differences in doing this with 'OCO' vs the 'Parent' property?

@Ed-Bartosh will this fix get merged into your main CCXT branch?

A trailing stop and a (normal) stop works completly different. For example on position bought at $10 a stop at say $9 would protect against a loss. A trailing stop of 20% would ensure that you sold if the price drops 20% at any time of the life span on the position. So if the price increased to $20 and drops to $16 it would be sold with a profit of $6.

A normal stop is static protecting against a loss and a trailingstop dynamic and typically primarily used to lock in a profit while keeping the upside of further increases.