You don't pass to the SimpleMovingAverage the ha_close line (self.hk.close) but rather a specific value of it (self.hk.close[0]).

If you look at the implementation of class HeikinAshi, it has line alias

    lines = ('ha_open', 'ha_high', 'ha_low', 'ha_close',)

    linealias = (
        ('ha_open', 'open',),
        ('ha_high', 'high',),
        ('ha_low', 'low',),
        ('ha_close', 'close',),
    )

The error has to do with division with zero, so I would suspect it has to do with the indicator. There is a parameter for RSI the safediv that you can set to True.

If you think that problem is with reading data, then I would suggest you check that bar(s) in next has the correct data

Hope this helps

Hi,

I am trying to change the rollover class to use it with input data feeds added in cerebro. I guess in this case next for the data feeds is also called by cerebro and thus no need to be called within rollover. I wonder how I can check for the end of the data feed without using the check for _next being False.

Many thanks

cerebro.setbroker(<instance of your broker class>)

Haven't tested it, but I think your csv file looks more like the following.

params = (
# ('nullvalue', float('NaN')),
('dtformat', '%Y-%m-%d %H:%M:%S'),
('datetime', 5),
('Open', 0),
('High', 1),
('Low', 2),
('Close', 3),
('Volume', 4),
)

Without having tested it, perhaps you can use in next of strategy B self.cerebro.runningstrats[self._id ^ 1] to get access to strategy A

I guess by filters you mean indicators and not data filters. This is an example having a list of indicators (in this case SimpleMovingAverage applied on close), one per data feed.

class TestStrategy(bt.Strategy):

    params = (
        ('maperiod', 15),
    )

    def __init__(self):
        self.smas = []
        for d in self.datas:
            self.smas.append(bt.indicators.SimpleMovingAverage(d.lines.close, period=self.params.maperiod))

I even like better the _nextforce = True pointed out to me by @backtrader some time ago.

class CustomIndicator(bt.Indicator):
    _nextforce = True
...

Another solution that does not use cheat on open. It is a cheat-on-close broker and you buy on open by disabling the coc and sell on close by enabling it.

class TestStrategy(bt.Strategy):

    def next(self):
        order1 = self.close(self.datas[0], coc=True)
        order2 = self.buy(self.datas[0], coc=False)

if __name__ == '__main__':
    cerebro = bt.Cerebro()
    cerebro.broker.set_coc(True)
    ....

The strategy can turn on/off a parameter of the indicator according to whether in trade or not. Have a look at the dynamic indicator example
https://www.backtrader.com/blog/posts/2018-02-06-dynamic-indicator/dynamic-indicator.html