Detect new resampled price

algoguy235

I have a code that uses both a tick data and a resampled data, let's call it 5 min period. I have a set of logics that only need to occur when a NEW 5m candle comes in. This will save a ton of computation. In order to detect this new price, here's what I've been doing:

(Lf = low frequency, meaning resampled data)

class St(bt.Strategy)
    def __init__(self):
        self.Lf_dt = self.datas[1].datetime
        # memeory functionality
        self.__previous_Lf_dt = None

    def next(self):
        # Check to see if there's a new Lf candle printed this period
        self.__new_Lf_print   = (self.Lf_dt[0] != self.__previous_Lf_dt)
        self.__previous_Lf_dt = self.Lf_dt[0]

        if self.__new_Lf_print:
            execute_Lf_logics()

Is there an easier way to do this? something built-in perhaps?

backtrader

Use the length of the data feed: len(data1). It will only change when new data is available.

algoguy235

Ahh yes that makes sense. Something like this.

class St(bt.Strategy)
    def __init__(self):
        # memeory functionality
        self.__previous_Lf_len = 0

    def next(self):

        if len(data1) != self.__previous_Lf_len:
            execute_Lf_logics()
            self.__previous_Lf_len = len(data1)

backtrader

In any case I would recommend against prefixing your member attributes with __. But it is your code of course.

algoguy235

Yes fair enough. But you would agree this is the most efficient way to accomplish this?

class St(bt.Strategy)
    def __init__(self):
        # memeory functionality
        self.previous_Lf_len = 0

    def next(self)
        if len(data1) != self.previous_Lf_len:
            self.new_Lf_print = True
            self.previous_Lf_len = len(data1)

        if self.new_Lf_print:
            execute_Lf_logics()

backtrader

Efficiency depends on your entire logic. If you simply want to act on a new bar, that's efficient enough for sure.

algoguy235

for reference - very similar issue here:

Strategy execution on different timeframes and multi-data

algoguy235

This is the most efficient solution I've found.

class St(bt.Strategy)
    def __init__(self):
        self.new_Lf_print = self.data1.datetime() != self.data1.datetime()(-1)

    def next(self)
        if self.new_Lf_print[0]:
            execute_Lf_logics()

By putting the empty brackets after the line object (datetime) it creates a linecoupler that is coupled to the base sample rate. Documentation here.

This slightly outperforms the len() calculation from the previous. I wasn't able to get len() as a line object in the init function, so datetime works well enough.

I also like this solution because the code is clean. Only requires a single line in init.

backtrader

@algoguy235 said in Detect new resampled price:

I wasn't able to get len() as a line object in the init function

len is a builtin and cannot be overridden. Given it is only meant to check a single value (because it increase monotonically) it seems pointless to have a replacement function for it to deliver a lines object.