@lampalork following @backtrader advise, i have tried the strategy selection pattern. it works great in my case. thanks! For ease of reference: https://www.backtrader.com/blog/posts/2016-10-29-strategy-selection/strategy-selection/

Wow, thanks for doing the hard work!

@backtrader I'll try this. Good input, thanks a lot

@tianjixuetu hi, as always, if you expect people to help you, you have to prepare a mini reproduceable example with code and data that highlights the problem you are facing. You can't expect people to try to guess what you are doing. This is not intended to be mean but rather to help you get the support you need

Dear All,

I know how to perform an optimization with backtrader. Let us say that I have a strategy with 2 parameters param1 and param2. The standard way of defining an optimization is done via (for instance)

cerebro.optstrategy(TradingTheCurveStrategy, param1 = range(1,10) , param2 = [3,6,9])

when calling optstrategy() will compute all the combination (product operation) of param1 and param2 and add them to the strategies to be backtested during the optimization process.

My question is as follows: Is there a simple way for me to specify exactly which combination of these 2 parameters I want to test. It can be useful when for instance, there is a relationship between these 2 parameters (some combination do not make sense) and we want to shorten the time of the optimization.

Thanks in advance

Kind regards
Lamp'

Hi @tianjixuetu, I have some logic in self.next() to chain these contracts. In other words, to know at any point of time what is the front contract (n=1) and all the other n contracts in the future.

yes, my bad. I meant this:

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

@lampalork said in Dynamic parameters passed to cerebro.optstrategy():

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

thanks very much, guys. I finally took the time to look at this and it works. Furthermore solution is much simpler than expected. A simple override of prenext() does the job! hopefully, I manage to reduce the memory footprint of my backtest (20 years on a basket of 12 commodities futures with individual contracts; you can imagine how many Datafeed I have to handle!)

    def prenext(self):
        current_date = bt.utils.date.num2date(self.datetime[0])
        if current_date >= self.params.start_date:
            self.next()

Hi, this is what I have done and indeed it works. Mysteries of pickle! thanks a lot

Dear Backtrader community,
I have implemented a strategy that uses the indicator bt.indicators.PercentRank(). When running optimization on this strategy, I get the following error message? Any idea/suggestions on how to go around this issue?

Thanks and regards,
Lamp'

multiprocessing.pool.MaybeEncodingError: Error sending result: '[<strategies.XXXStrategy object at 0x000002D1A61754A8>]'. Reason: 'PicklingError("Can't pickle <function PercentRank.<lambda> at 0x000002D1A716A510>: attribute lookup PercentRank.<lambda> on backtrader.indicators.percentrank failed")'