Rerun backtests with different start dates

brettelliot

Hi,

I'm testing asset allocations using decades of monthly data and I want to make sure the starting month doesn't have an outsized impact on the results. Is there a way to do that?

The best idea i have so far is to create a strategy that takes a start date as a parameter and use optstrategy to pass that in for every different start date I want to test. Then the algo just wouldnt do anything before the start date. However that impacts the analyzers which consider the start date by the start of the data.

Thoughts?

Thanks,
Brett

davidavr

@brettelliot I would think the easiest way is to just pass start and end date arguments to data feed calls for each data set.

data = btfeeds.GenericCSVData(
    dataname=filename,
    fromdate=start_date,
    todate=end_date,
    ...
)

brettelliot

@davidavr I planned on using optstrategy which means the data feeds are loaded once before running the backtests.

I guess i could skip using optstrategy and build my own loop that runs a single threaded backtest for each month. I might lose multithreads but I also don't have to adjust analyzers.

Ill give it a shot!

run-out

@brettelliot You can use python multiprocessing to run seperate backtest.