resampling - first resampled bar is sampled incorrectly

Jens Halsberghe

I'm trying to resample from 1 minute to 2, 5 and 10 minutes. The only problem I'm having is when resampling, the first OHLC data I'm logging of the 2,5 and 10 minute is exactly the same as the first 1 minute OHLC data. after that it resamples properly, however it then is out of sync.

an example will make it more clear when sampling 1 minute to 2 minute data:

Output in the log:

Starting Portfolio Value: 1000000.00
2020-08-12 00:00, O: 1.17383, H: 1.17391, L: 1.17372, C: 1.17391
2020-08-12 00:02, O: 1.17391, H: 1.17408, L: 1.17384, C: 1.17384
2020-08-12 00:04, O: 1.17384, H: 1.17384, L: 1.17377, C: 1.17377
2020-08-12 00:06, O: 1.17377, H: 1.17389, L: 1.17376, C: 1.17386

dataframe info:

                        Open    High    Low    Close   Volume
Datetime					
2020-08-12 00:00:00	1.17383	1.17391	1.17372	1.17391	23
2020-08-12 00:01:00	1.17391	1.17408	1.17390	1.17390	18
2020-08-12 00:02:00	1.17390	1.17390	1.17384	1.17384	33
2020-08-12 00:03:00	1.17384	1.17384	1.17378	1.17378	22
2020-08-12 00:04:00	1.17378	1.17378	1.17377	1.17377	25
2020-08-12 00:05:00	1.17377	1.17382	1.17376	1.17382	39
2020-08-12 00:06:00	1.17382	1.17389	1.17382	1.17386	36

The first candle should be O:1.7383 H:1.17408 L:1.17372 C: 1.17390 but for some reason it always takes over the very first minute candle which causes all the rest of the candles to be out of sync.

my code below:

class Resampling_test(bt.Strategy):
 
    def log(self, txt, dt=None):
        ''' Logging function for this strategy'''
        dt = dt or self.datas[0].datetime.datetime(0)
        print('%s, %s' % (dt.strftime("%Y-%m-%d %H:%M"), txt))
    
    def __init__(self):
        
        pass
       
    def next(self):
        
        self.log('O: %.5f, H: %.5f, L: %.5f, C: %.5f' %(self.datas[0].open[0], self.datas[0].high[0], self.datas[0].low[0], self.datas[0].close[0]))

if __name__ == '__main__':

    cerebro = bt.Cerebro()

    # Add a strategy
    cerebro.addstrategy(Resampling_test)
    
    # Create a Data Feed
    data = bt.feeds.PandasData(dataname=df2020)
    two_minute = cerebro.resampledata(data, timeframe=bt.TimeFrame.Minutes, compression=2)
    
    # Print out the starting conditions
    print('Starting Portfolio Value: %.2f' % cerebro.broker.getvalue())
    
    cerebro.run()
    
    # Print out the final result
    print('Final Portfolio Value: %.2f' % cerebro.broker.getvalue())

many thanks

Jens Halsberghe

I've been able to fix the bars by using

boundoff=-1

for the 2 minute timeframe and

boundoff=1

for the 5 and 10 minute timeframe

That's great but now the time is messed up

Starting Portfolio Value: 10000.00
2020-08-12 00:01, O: 1.17383, H: 1.17408, L: 1.17372, C: 1.17390
2020-08-12 00:03, O: 1.17390, H: 1.17390, L: 1.17378, C: 1.17378
2020-08-12 00:05, O: 1.17378, H: 1.17382, L: 1.17376, C: 1.17382
2020-08-12 00:07, O: 1.17382, H: 1.17389, L: 1.17382, C: 1.17386

it should be even starting from 00:00

I've read the following about boundoff from the documentation Here:

If for example the resampling is from 1 minute to 15 minutes, the default behavior is to take the 1-minute bars from 00:01:00 until 00:15:00 to produce a 15-minutes replayed/resampled bar.

I'm not understanding this very well as in my mind it makes sense to take 1 minute bars from 0:00 to 14:00. so not sure why the default behavior is like this. regardless, it can be fixed using boundoff but then it messes up the datetime.

dasch

you could also play with rightedge=True and boundoff=x

see this thread: https://community.backtrader.com/topic/970/backtest-vs-live-bars-off-by-1-discrepancy

Jens Halsberghe

@dasch thanks for your response. yes as described in my previous post, I have used boundoff which fixes the bar issue. when I use rightedge=True or any of the other parameters. it doesn't fix my timebar issue however

dasch

@Jens-Halsberghe yes, i saw your solution. That’s when I remembered the other thread.

Jens Halsberghe

@dasch I just read the full thread. this seems to be an outstanding issue for a while. I hope it'll get fixed at some time. I really like the library and I want to use it for livetrading but I hope this problem doesn't jeopardize it. what do you now use for livetrading?

dasch

I am still using backtrader. For simple stuff I write stuff directly using the oanda api.