Resample data with timeframe.days at specific (non-midnight) time

algoguy235

I'm interested in using an indicator that uses daily candles, not from 0:00a to 11:59p, but rather from 4:16p to 4:15p. Can I do this using the cerebro.resampledata(), perhaps with the boundoff parameter?

If not, is there another way?

backtrader

Set the sessionend parameter of the data feed which is going to be resampled. It is a parameter for all data feeds, because it is inherited from the superclass.

See Docs - Data Feeds Reference

algoguy235

That worked perfectly, thanks @backtrader.

It wasn't clear how sessionend and sessionstart parameters are used in other places in the engine. My next challenge is that I want to start the session at 600p and end at 415p. setting the sessionstart param did not solve this, My daily candles are still going form 415 to 415.

backtrader

@algoguy235 said in Resample data with timeframe.days at specific (non-midnight) time:

My daily candles are still going form 415 to 415

The best thing would be that you show what you do and how you come to the conclusion that the daily candle goes from 04:15 to 04:15.

In any case your problem is probably the resolution of your timeframe. A sessionstart larger than a sessionend makes only sense when you have a sub-day timeframe. Else it would be a running-backwards session.

algoguy235

@backtrader

Is it possible to change the sessionend parameter in cerebro.resampledata?

    cerebro = bt.Cerebro()
    cerebro.addstrategy(strat)

    data = btfeeds.GenericCSVData( # 1-min csv feed 
                    dataname     = datapath,
                    fromdate     = startDate_feed,
                    todate       = endDate,
                    nullvalue    = 0.0,
                    openinterest = -1,
                    timeframe    = bt.TimeFrame.Minutes,
                    sessionend   = session_End,
                    dtformat     = ('%m/%d/%Y'),
                    tmformat     = ('%H:%M'),
                    time=1, open=2, high=3, low=4, close=5, volume=6)

    # Add the 1m Data Feed to Cerebro
    cerebro.adddata(     data, name = 'a0_Hf') 

    # create a daily data feed with standard session end time (session_End)
    cerebro.resampledata(data, name = 'a0_d1a', timeframe = bt.TimeFrame.Days, compression = 1)

    # create a daily data feed with a different session end time
    cerebro.resampledata(data, name = 'a0_d1b', timeframe = bt.TimeFrame.Days, compression = 1, sessionend = datetime.time(14,15))

I want two 1-day resampled feeds with different sessionend. One idea I had was to feed the CSV data in twice with the different sessionend. I think that would work, but I was hoping I could do this using only resample data. Thoughts?

It seems the whole platform takes the sessionend parameter from the first feed, and ignores it later on, even if I leave it empty/default.

backtrader

A resampled version of a data feed must have the same sessionend as the original.

Li Mike

@backtrader In chinese future market, sessionstart is 21:00 and sessionend is 15:00 next day. So, you see sessionstart is indeed larger than session end. Can backtrader process this situation correctly?