Cannot access the resampled data in self defined indicators

yoghurtshawn

I had a data source in Minutes time frame, and I resampled it into 15 minutes data. However, I cannot get the resample data value in my custom Indicators' __init__ function (it calls some talib functions). It says "Exception: inputs are all NaN". Is there anything wrong?

backtrader

@yoghurtshawn said in Cannot access the resampled data in self defined indicators:

Is there anything wrong?

Knowledge osmosis is not yet possible. No data, no code, no nothing ... no answer

yoghurtshawn

Sorry.

The data look like this, being 1MIN kline and have been added.

Then, I resample it with cerebro.resample(data, timeframe=bt.TimeFrame.Minutes, compression=15), hoping to get klines of 15MIN.

It can be accessed in bt.indicators.SimpleMovingAverage(self.datas[1].close, period=30) and it works fine.

However, when the indicator is defined by myself, for example,

class SMACustIndicator(Indicator):
    lines = ('sma_cust',)
    params = (
        ('period', 30),
    )
    plotinfo = dict(subplot=False)

    def __init__(self):
        self.sma_cust = talib.SMA(
            pd.Series(self.data.close.array),
            timeperiod=self.params.period
        )

    def next(self):
        cur = len(self)
        self.lines.sma_cust[0] = self.sma_cust.iloc[cur-1, ]

It raises "Exception: inputs are all NaN".

backtrader

@yoghurtshawn said in Cannot access the resampled data in self defined indicators:

    def __init__(self):
        self.sma_cust = talib.SMA(
            pd.Series(self.data.close.array),
            timeperiod=self.params.period
        )

At that stage there is no data available, which means you are creating a pd.Series full of NaN, hence the result.

ta-lib is integrated with backtrader, use the integration directly. Doc - Indicators - ta-lib

yoghurtshawn

Thanks! @backtrader

One more question, do you mean, the data will only be filled at next() calls?

backtrader

No, I mean that you are initializing a pd.Series with an array which is full of NaN values.