Backtest Granularity
Hi,
Am attempting to achieve the granularity suggested in the samples. However, it appears that what is actually created is more frequent sampling, sure, but on the indicator series generated by the faster tick stream...
How do I achieve the following:
I have an indicator for a strategy (primary time series) which when "live" is checked (and triggered) on a tick by tick basis related to the primary time series (ie tick by tick I see and trade the primary time series paint).
Of course backtesting I get triggers only based on the primary time series (ie late).
I want to improve the granularity for backtest puposes. So I set up a secondary (1 minute) time series as advertised in the sample and literature, I do indeed get 1 minute triggers, but only on the indicator generated by the secondary (1 minute) time series. So if I run a backtest or optimise, I get the same triggers for 60 minute as 2 minute primary bars.
Is there a way to replicate the "live" performance (albeit on 1 minute bars, which will suffice), but using the primary bar paint?
To give an example: if you have, say a 60 minute SMA [4], you will see the SMA of the last four hourly bars, but if "OnBarUpdate" is false, you will see that 60 minute SMA painting live on the current, unfinished bar. If you want to backtest of the 60 minute SMA[4], but as it would be painting live (well, every 1 minute), how do you do it?
Rgds,
Am attempting to achieve the granularity suggested in the samples. However, it appears that what is actually created is more frequent sampling, sure, but on the indicator series generated by the faster tick stream...
How do I achieve the following:
I have an indicator for a strategy (primary time series) which when "live" is checked (and triggered) on a tick by tick basis related to the primary time series (ie tick by tick I see and trade the primary time series paint).
Of course backtesting I get triggers only based on the primary time series (ie late).
I want to improve the granularity for backtest puposes. So I set up a secondary (1 minute) time series as advertised in the sample and literature, I do indeed get 1 minute triggers, but only on the indicator generated by the secondary (1 minute) time series. So if I run a backtest or optimise, I get the same triggers for 60 minute as 2 minute primary bars.
Is there a way to replicate the "live" performance (albeit on 1 minute bars, which will suffice), but using the primary bar paint?
To give an example: if you have, say a 60 minute SMA [4], you will see the SMA of the last four hourly bars, but if "OnBarUpdate" is false, you will see that 60 minute SMA painting live on the current, unfinished bar. If you want to backtest of the 60 minute SMA[4], but as it would be painting live (well, every 1 minute), how do you do it?
Rgds,
Comment