How to use indicator with multiple outputs?

[edtechre]

How to use indicator with multiple outputs?

[edtechre]

Hi @none2003,

I may be able to add support for this by allowing multiple indicator names:
pybroker.indicator(name=['upperband', 'middleband', 'lowerband'], lambda data: talib.BBANDS(data.close))

I will look into the feasibility of this.

[edtechre]

Hi @none2003,

Yes, this is because indicators only support one output. You will need to create separate indicators for each of the upperband, middleband, and lowerband outputs.

[none2003]

Hi @edtechre,

I have used backtesting.py, which also uses a decorator to encapsulate the results returned by the indicator:

self.upperband, self.middleband, self.lowerband = self.I(talib.BBANDS, self.data.Close)

These return values can then be used in the "next" method (similar to the "execute(ctx: ExecContext)" method in pyb):

...
if self.data.Close[-1] > self.upperband[-1]:
    pass
...

This is much easier than defining a method for each return value of the indicator, and I really hope pyb supports such a feature.

Maybe you can take a look at their implementation of this "I" method:

https://github.com/kernc/backtesting.py/blob/master/backtesting/backtesting.py line 78

[none2003]

Hi @none2003,

I may be able to add support for this by allowing multiple indicator names: pybroker.indicator(name=['upperband', 'middleband', 'lowerband'], lambda data: talib.BBANDS(data.close))

I will look into the feasibility of this.

Hi @edtechre,

This looks good, have you had time to research on this?

[edtechre]

Hi @none2003,

I have decided not to make this change as it would not be compatible with how indicator calculations are parallelized across multiple CPUs. My recommendation is to create separate indicators for each output, which is only a minor inconvenience.

[tsunamilx]

Hi @edtechre

So what you suggest is like:

bbup = pybroker.indicator('bbup', lambda data: talib.BBANDS(data.close).upperband)
bbdn = pybroker.indicator('bbdn', lambda data: talib.BBANDS(data.close).lowerband)
bbmd = pybroker.indicator('bbmd', lambda data: talib.BBANDS(data.close).middleband)

Is that right? If so, will it repeat the bbands calculation 3 times?

[edtechre]

Hi @tsunamilx,

Yes, that is true. But the TA-Lib calculation is very fast, so you should see very little difference in performance.