How to use indicators that return a Dataframe / Series?

[valex91]

How to use indicators that return a Dataframe / Series?

[xmatthias]

You will also need to be rather carefull with this - this indicator is using the FULL timerange you're passing in - so in backtesting, it's looking at the whole range and has therefore a lookahead bias (it'll know at all times what the last point in the series was).

You'll need to use it carefully - and apply it in a loop of "smallish" steps to get this right ...

[xmatthias]

i'm ... not quite sure what you are using (and how).
Are you able to share a bit more of the implementation you're using? maybe this will allow us to help you better.

[valex91]

I'm attempting to use:

technical/technical/trendline.py

Line 8 in a1e92f3

def gentrends(dataframe, field='close', window=1 / 3.0, charts=False):

and

technical/technical/indicators/indicators.py

Line 734 in a1e92f3

def fibonacci_retracements(df, field='close') -> DataFrame:

for the trendline indicator I'm attempting to use it in the following manner:

    def populate_indicators(self, dataframe: DataFrame, metadata: dict) -> DataFrame:
        dataframe['rsi'] = ta.RSI(dataframe, timeperiod=9)
        dataframe_long = resample_to_interval(dataframe, 30)

        # Bollinger Bands
        bollinger3std = qtpylib.bollinger_bands(qtpylib.typical_price(dataframe), window=20, stds=3)
        dataframe['bb3std_lowerband'] = bollinger3std['lower']
        dataframe['bb3std_middleband'] = bollinger3std['mid']
        dataframe['bb3std_upperband'] = bollinger3std['upper']

        # trendline
        trend = segtrends(dataframe_long, field='close')
        dataframe_long['trend'] = trend
        # Combine the 2 dataframes
        dataframe = resampled_merge(dataframe, dataframe_long, fill_na=True)

        return dataframe

and to populate the buy indicator:

 def populate_buy_trend(self, dataframe: DataFrame) -> DataFrame:
        dataframe.loc[
            (
                    self.good_deal_buy_signal(dataframe) & (
                    dataframe['close'] > dataframe['resample_30_trend']['Min Line'])), 'buy'] = 1

        return dataframe

but also tried to populate the indicator in the following manner as a dataframe is returned:

  trend = segtrends(dataframe_long, field='close')
        result = dataframe.merge(trend, how='cross')
        
        return result

but then the script seems to hang, I suspect because the merge is quite heavy on performance.

[xmatthias]

a merge is (usually) not heavy on performance - unless you use rather complex conditions.

I've never seen a cross merge though, and i'm not sure that'll work for timeseries.

that result can be rather big (with probably limited benefits, considering that it will combine every row with every other row - i also think that this will lead to a lookahead bias, considering how cross merges work (and you can't really use it anyway as it'll multiply your rows).

What's good_deal_buy_signal() do? it must return a boolean series (of the correct length!), otherwise this will fail.

[valex91]

this is the good_deal_buy function:

    def good_deal_buy_signal(self, dataframe):
        return dataframe['close'] < dataframe['bb3std_lowerband']

but that alone is working as it should, the fact that I'm trying to figure out is for those indicator that return a dataframe of their own how do you pick the relevant row which hold the information to the row that I'm looking into.

In pseudo code what I'm trying to do is
trend_lines = segtrends(dataframe_long, field='close') dataframe['support'] = trend_lines['Min Line'] dataframe['resistence'] = trend_lines['Max Line']

to then use that information as a guard in the buy signal as
dataframe['close'] > dataframe['support']

[xmatthias]

well - this is not pseudo code, but code how you should actually use it 😆

[valex91]

with that code the result of print(dataframe.tail()) is:

with the following populate_indicators function

   def populate_indicators(self, dataframe: DataFrame, metadata: dict) -> DataFrame:
        dataframe['rsi'] = ta.RSI(dataframe, timeperiod=9)
        dataframe_long = resample_to_interval(dataframe, 30)

        # Bollinger Bands
        bollinger3std = qtpylib.bollinger_bands(qtpylib.typical_price(dataframe), window=20, stds=3)
        dataframe['bb3std_lowerband'] = bollinger3std['lower']
        dataframe['bb3std_middleband'] = bollinger3std['mid']
        dataframe['bb3std_upperband'] = bollinger3std['upper']

        trend_lines = segtrends(dataframe_long, field='close')
        dataframe_long['support'] = trend_lines['Min Line']
        dataframe_long['resistence'] = trend_lines['Max Line']

        dataframe = resampled_merge(dataframe, dataframe_long, fill_na=True)

        return dataframe

[xmatthias]

i never used trendlines / seglines - maybe it's not supposed to "always" give you value?

try it in a "simple" setup, maybe? without resampling, directly plotting the "dataframe_long" dataframe.

[ronmegini]

with that code the result of print(dataframe.tail()) is:

with the following populate_indicators function

   def populate_indicators(self, dataframe: DataFrame, metadata: dict) -> DataFrame:
        dataframe['rsi'] = ta.RSI(dataframe, timeperiod=9)
        dataframe_long = resample_to_interval(dataframe, 30)

        # Bollinger Bands
        bollinger3std = qtpylib.bollinger_bands(qtpylib.typical_price(dataframe), window=20, stds=3)
        dataframe['bb3std_lowerband'] = bollinger3std['lower']
        dataframe['bb3std_middleband'] = bollinger3std['mid']
        dataframe['bb3std_upperband'] = bollinger3std['upper']

        trend_lines = segtrends(dataframe_long, field='close')
        dataframe_long['support'] = trend_lines['Min Line']
        dataframe_long['resistence'] = trend_lines['Max Line']

        dataframe = resampled_merge(dataframe, dataframe_long, fill_na=True)

        return dataframe

Hey, how are you print the indicators dataframe table?

[xmatthias]

well - print(dataframe.tail()) (as specified above).

I'll now also close this issue as i think the topic has been resolved.