Buy and hold strategy for daily timeframe

Blueshift
1

Default behaviour of buy and hold strategy is executed on minute level data.

Is there exaple how to do it on a daily data ? 

(This confusion comes mainly after new interface)

2

Compared to running backtesting on daily datasets, minute data usually is more accurate at the cost of a slight increase in time to run. Can you please share your use case where you require daily data? Also see here for an interesting read on why low-frequency data (such as daily data) for backtesting can be misleading. At present, we support only minute data after Blueshift 2.0. This has extended the universe (bigger than earlier minute dataset, and even daily dataset), but has a shorter history than earlier daily dataset (we could not yet locate a reliable source of precise minute data before the available period, and in backtesting, garbage-in/garbage-out is dangerous).

3

Thanks for the response 

I had performed backtest on trading strategies , which operate on daily data, (obtained data using broker APIs)

I believe some strategies are happy with daily data, however these need a broad period (with different phases of market) for backtests 

Tough minute level data can be upsampled in the program (I think API provides an option to get daily data as well), the lmiitation of only 3 years data hampers my test coverage



I think one of the below solutions would suffice me 

  • Option to use custom data set for backtesting
  • Longer history availablity 
While I understand that providing minute level for a larger history need a very careful consideration to accuracy, it would suffice many like me , to have a custom data set.

Thanks
Prasanna

 
4

Ah yes, I think this is an issue for the users of the NSE dataset. The cleanest solution will be for us to extend the history. A lot of efforts go on behind the scene at Blueshift to ensure data quality. For example, check out IOC share price on any other platform between 2020 and 2021. You will see it shows a flat performance. This is not the case actually, as IOC paid significant amount of dividends during this perod. Most other platforms will exclude such effects and will present a wrong outcome. This is the risk of using custom datasets. If they are not properly prepared, it can be very misleading.



Having said that, we do have a plan to support custom dataset. The primary goal is to support non-standard (or alternative) datasets, not exactly pricing data. But I guess that should work for pricing data as well - although I will not recommed it unless you have properly prepared it. At present, the ETA for this feature to go live is not decided.