Portfolio Run with as much data series as possible

Bo

Hi Community,

I am backtesting a simple momentun strategy on a set of futures...however, these futures are listed on different start time. To be a bit more concrete, Assuming I have:

Future1: from 2018 to 2021
Future2: from 2019 to 2021
Future3: from 2019 to 2020

I would like to apply a simply momentun strategy on All three Futures. During 2018-2019, all initial cash value will be allocated to Future 1. During 2019-2020, available cash will be allocate equally to Future 1, 2 and 3. During 2020-2021, value will be allocate to Future1 and 2.

I know I can run these futures separately and aggregate the results. However, this would be super cumbersome when it comes to aggregation and more importantly not very accurate since profit/loss could be rebalanced in portfolio run but not if I ran them seperately.

But, to kick off portfolio run, I am facing a problem which is, based on backtrader, the actually backtesting would start only when Future2 is fully loaded, i.e. the first buy signal execution can only happen after 2019. And I am also not sure what is going to happen after 2020 since Future3 no longer exsits.

I have tried to load the data by specifying fromdate and todate like follow:

*data1=data.iloc[-100:,:]
data1.index=data1.DATETIME

data2=data.iloc[-50:,:]
data2.index=data2.DATETIME

cerebro = bt.Cerebro() # create a "Cerebro" engine instance
datax1=bt.feeds.PandasData(dataname=data1,fromdate=min(data1.DATETIME),todate=min(data1.DATETIME),timeframe=bt.TimeFrame.Days)
datax2=bt.feeds.PandasData(dataname=data2,fromdate=min(data2.DATETIME),todate=min(data2.DATETIME),timeframe=bt.TimeFrame.Days)

cerebro.broker.setcash(1000.0)
cerebro.broker.set_slippage_perc(0.0001)
cerebro.broker.setcommission(commission=0.0001)
cerebro.adddata(datax1) # Add the data feed
cerebro.adddata(datax2) # Add the data feed
cerebro.addstrategy(Naive_Momentun_Print) # Add the trading strategy
cerebro.broker.addcommissioninfo(CommInfoFractional())
cerebro.run()*

But it gives me this error:
File "/lib/python3.7/site-packages/backtrader/linebuffer.py", line 672, in once
dst[i] = src[i + ago]

IndexError: array assignment index out of range

I am not sure what to try out next. Hope someone in the community can help out.

Thanks in Advance!
Bo

####################################
#Attached below is my native stategy:
####################################
class Naive_Momentun_Print(bt.Strategy):
# list of parameters which are configurable for the strategy
params = (
# Standard Parameters
('period', 25),
)

def __init__(self):
    self.mom = [bt.indicators.MomentumOscillator(i, period=self.params.period) for i in self.datas]
    
    
def notify_order(self, order):
    if order.status in [order.Completed]:
        if order.isbuy():
            self.log('Buy Order EXECUTED, %.2f, SIZE of: %.2f, Asset of: %s' % (order.executed.price,order.executed.size,order.info))
        if order.issell():
            self.log('Sell Order EXECUTED, %.2f, SIZE of: %.2f, Asset of: %s' % (order.executed.price,order.executed.size,order.info))


def notify_trade(self, trade):
    if not trade.isclosed:
        return
    self.log('OPERATION PROFIT, GROSS %.2f, NET %.2f' %
             (trade.pnl, trade.pnlcomm))

def start(self):    
    self.order = None  # sentinel to avoid operrations on pending order

def next(self):
    print (""""Time: {0}""".format(self.datas[0].datetime.datetime(0)))
    
    c = [i.momosc[0] for i in self.mom]
    
    for i in range(0,len(c)):
           if self.broker.getposition(data=self.datas[i]).size==0:
               self.order_target_percent(data=self.datas[i], target=0.49)
           else:
               self.order_target_percent(data=self.datas[i], target=0)

def log(self, txt, dt=None):
    dt = dt or self.datas[0].datetime.datetime(0)
    print('%s, %s' % (dt.isoformat(), txt))

run-out

@bo This answer gives a rough idea of how to approach it. Create an index covering the entire period. Use pandas to reindex your data for the whole period using 0 value when the future is not trading. Load into backtrader using pandas feed.

Then filter out 0 datas in your next

Bo

@run-out IT WORKS!
Thank you for your swift prompt. My struggling was having passed unaligned datetime index. Once the index are the same, I can do all kinds of filtering and calculation.
Thanks again!

Bo

@run-out Hi

I am running into another issue using your method. If I algined datatime, then when I calculate indicators like MACD, the futures that issued later in time will all be NAN. I won't be able to leverage powerful indicator functions build in Backtrader.

If I filled all nan close price to 0, then I won't be able to calculate MACD currectly in the first data points.

Is there any other elegate solutions? such as specify min_period or have indicator calucation ignoring nan?

vladisld

@bo Another option is to utilize the prenext method. See the example here:
https://www.backtrader.com/blog/2019-05-20-momentum-strategy/momentum-strategy/#next-and-prenext