ok, found what I was looking for. For the benefits of others...

d = {'period': [14, 16, 20 , 22, 25, 30], 'period2': [33, 39]}
cerebro.addstrategy(MyHolyGrailStrategy, *[], **d)

cheers

Dear All,
I'm familiar with the following syntax:

cerebro.optstrategy(MyHolyGrailStrategy, period=range(1, 20), )

But how can I call optstrategy when the parameters to be optimized and their respective ranges are only known at runtime?
For addstrategy(), I'm using something like

cerebro.addstrategy(MyHolyGrailStrategy, parameters)

where parameters is a dictionnary that is received by the _init method of my strategy

def __init__(self, params):

But the same approach does not seem to be working here.

Many thanks in advance
Kind regards,
Lamp'

@backtrader noted. thanks

Thanks, ab_trader and sorry for getting back so late. I'll double-check your claim and will update if it does not behave like you describe; for now, my trick works for it feels unnecessary.

Dear Backtrader community,

how would you implement different slippage for differents symbols/datafeed. From what I can see (i have also tested it to be sure), the function cerebro.broker.set_slippage_perc() does not take a datafeed name unlike the function cerebro.broker.setcommission()

many thanks in advance
Kind regards

yes, for other strategies. but in this case, I need to be able to trade at different point of the forward curve (e.g. front contract and contract expiring in one year; a few weeks later front contract and contract expiring in 6 months). Plus in this case, it is actually useful to me to know which actual individual contract has been traded

thanks, I'll take a look. Profiling the backtest is any case probably a good start

Dear All,

I'm currently backtesting a strategy that needs a lot of datafeeds. If I understand correctly how backtrader works, one of the first step cerebro will perform (when doing cerebro.run()) is to pre-load the data (assuming preload = True which is the default setting). This step takes quite a bit of time in my case (maybe 3/4 minutes) but running through the bars afterwards is relatively fast (as fast as python can be:)); is there a way I could cache/store things so that i don't need to wait that time (3/4 minutes) everytime i run a backtest? I was thinking about pickling the cerebro object (i know cerebro can be pickled) but i do not know which cerebro's method to launch to load the data without launching cerebro.run(). or would you any other suggestion to shorten this annoying waiting time?

thanks and regards
Lamp'

Dear All,

I'm currently backtesting a Futures strategy that trades the entire curve (i.e. does not only trade front contract). It means that I need to work actual Futures contract rather than stitched futures contract (e.g. #1, #2, #3).
What I do is to add every Futures contract as a data feed and then the strategy has code to retrieve the datafeeds (that corresponds to contract that have not expired) sorted by expiration date. One of the issue I have faced is that these individual contracts don't have all data for every date (let assume I perform a backtest over 10Y; it is easy to understand that Dec2019 contract had no data early 2010) and therefore the backtest is only performed with a subset of the contracts (i.e. Datafeed) and only from the date where the last datafeed added to Cerebro has data.

I have temporary "resolved" this issue by reindexing my dataframes (and therefore adding nan date) before loading them to backtrader to ensure that all datafeeds have records from the start date till the end date I want. Is there by chance something more elegant than this I could do?

Many thanks
Regards,
Lamp'

Dear All,

I have a custom indicator that just calculates the ratio between 2 data series (data0 / data1). See indicator code and initialization below. I would like to plot the ratio indicator without ploting the 2 underlying series (i deactivate plotting for these series using data0.plotinfo.plot = False)? When i deactivate any of these data series, the ratio indicator automatically disappears. I think it is because the master relationship with underlying data. What is the best way to work around this? Thanks!

class DivisionIndicator(bt.Indicator):

    _mindatas = 2  # requires two (2) data sources
    alias = ('DVD',)
    lines = ('div',)

    params = (
        ('not_required', 20),
    )

    plotlines = dict(
        divisor_ind=dict(
            _name='DVD',
            color='green',
            alpha=0.50
        )
    )

    def nextstart(self):
        self.l.div[0] = self.data0[0] / self.data1[0]  # seed value

    def next(self):
        d = self.data0[0] / self.data1[0]
        self.l.div[0] = d if d else self.l.div[-1]

    def oncestart(self, start, end):
        self.line.array[start] = (
            self.data0.array[start] - self.data1.array[start])

    def once(self, start, end):
        d0array = self.data0.array
        d1array = self.data1.array
        larray = self.line.array

        prev = larray[start - 1]
        for i in range(start, end):
            d = d0array[i] / d1array[i]
            larray[i] = prev = d if d else prev

class CalvadosStrategy(bt.Strategy):
      def __init__(self):
        my_ratio = DivisionIndicator(self.datas[1], self.datas[2])
        ...