Backtesting problem??

dummydost
dummydost edited October 2016 in General
So, while streaming live data, we get data with an interval of 700 ms - 1 ms (for nifty stocks, liquid stocks, that is) but historical API provides only minimum 1 minute data range ( I maybe wrong!).

Wouldn't the backtesting results (for intra-day trading strategy with positions holding period ranging from a few minutes to 0.5-1 hour) be incorrect in this case, since the prices can fluctuate a great deal in on a minutely basis? How to deal with this?

Can someone please clarify?
  • rahulchitale
    There is significant variation during the volatile opening session within a 60 second period. I had done a bit of analysis and the swing was averaging around 0.5% intra-minute during the 2nd half after the initial circus has ended - it is more during the opening session. As you surmised, analysis isn't easy because the candles API have a granularity of 1 minute and intra day needs a lot more insight. Perhaps somebody from Zerodha can comment if they are planning to make it easier for such analysis efforts by including sub-minute data.
  • dummydost
    dummydost edited October 2016
    @rahulchitale If they provide sub-minute data, backtesting would actually start making sense, because we are talking about intraday strategies wherein positions may last only a few seconds to a maximum of 0.5-1 hour.
    Thanks for your input.
  • nithin
    Historical data is today typically saved as 1 minute OHLC (Open, High, Low, Close), this is the same with most data vendors as well (even in the developed markets). I know TR (Thomson Reuters) provides you historical tick data (1 tick a second kinds), but it costs a bomb.
    But I am adding your request on our list of things to do, see if it is viable to give sub 1 min OHLC data.
  • dummydost
    dummydost edited October 2016
    @nithin Tick data definitely costs a bomb :) , but we only need that much data as we see/receive while trading live (live stream). It'd make backtesting more accurate.
    Thank you for adding it to your to-do list.
Sign In or Register to comment.