Decreasing accuracy of pro backtest results in longer time frames?

• 12/09/2020 at 2:19 PM #153089
  Kovit

  Hi guys,

  Hope everyone is well and looking forward to the Christmas break.

  I was just wondering if the backtest results using longer/larger time frames would be less accuracte than the shorter/smaller time frames?

  If prorealtime takes a middle point of each candle when making backtest calculations it would be logical to assume a 1min candle has a smaller range and therefore be more accurate than a 15 minute or 1 hour candle ?

   

  

  Kovit

  Participant

  Topics: 25

  Replies: 84

  Been thanked: 13 times
  12/11/2020 at 6:12 AM #153305
  Finning

  Interesting that you mention that.

  I just saw this in one of my trials, weekly bars, AUDUSD mini.

  A system exit at 1.01977 – but the low of the candle is said to be at 1.04217

  So that means that the trade exited 224 pips below the low of the bar – where there is no price data for the candle.

  Not sure how that can happen?

  Tick by tick mode selected.

  Timeframe set to market timeframe UTC+0.

  

  Finning

  Participant

  Topics: 16

  Replies: 98

  Been thanked: 7 times
  12/11/2020 at 9:51 AM #153329
  GraHal

  Finning wrote:

  Not sure how that can happen?

  If on Real Live you need to query this with IG / your broker?

  

  GraHal

  Participant

  Topics: 204

  Replies: 8399

  Been thanked: 2041 times
  12/12/2020 at 11:59 AM #153485
  PeterSt

  Hi, I am not sure whether this is even less worth than 2c :

  First off, the graphs between direct PRT and PRT-IG are more not the same than the same.
  Next, let’s take into account that a data feed is as far as I know not coming from a broker, but from a third party (many of them, I’d say).

  The above would mean that PRT may get their data from (again if you ask me) the same data provider in both cases – IB and IG (hence PRT direct vs PRT-IG).

  This again would imply that back test results are the same in Direct vs IG which …
  implies that what IG makes of it in practice, is different. Thus, IG back testing would not represent reality for IG.
  On a side note, I am not claiming that in Direct PRT the results WILL represent reality. But what is and remains crucial is that reality in Direct vs IG is not equal*.

  *) Which already may spring from IG working with CFD’s ?

  Disclaimer : My text above could be complete BS, but I always wondered why my 1000s of back tests in IG with results you don’t want to know, fail on the very first trade, and so much so that it is not bearable.
  And no, I don’t over-optimize.

  0c,
  Peter

  

  PeterSt

  Participant

  Topics: 70

  Replies: 2287

  Been thanked: 695 times
  12/12/2020 at 12:19 PM #153488
  PeterSt

  Hence …
  I tried to align these two (which is difficult enough regarding the difference) and as you can see, it looks as if I selected two different Forex pairs. Still they are both EUR-USD (not Mini).
  Mind the 1 second bar length.

  Also, it is not clear to me whether you are already Live with PRT-IG V11, but this one is live (thus no possible fake-demo).

  And thus … how can we ever Back Test on this, when PRT would use different historical data than IG’s practice.
  Maybe others have more insight ? (I hope this is not off-topic)

  PS: I have given up fighting this long ago, but for these reasons I have also given up Auto Trading with IG.

  

  PeterSt

  Participant

  Topics: 70

  Replies: 2287

  Been thanked: 695 times
• Author
  Posts