ATR unexpected result

[GKonheiser]

I have created a dataseries (hCloses) in my multi Series strategy ( 10min (primary) and 1 Tick(secondary)). The dataSeries is the same as Cloeses[0] apart from the Value[0] of the data series that is set to Closes[1]:-

if(BarsInProgress == 0)
				{
					hClose.Set(1,Close[1]);
				}
				
				if(BarsInProgress == 1 && CurrentBars[baseBars] > 1)
				{
					hClose.Set(Closes[1][0]);
                                }

I have verified the that both series are as expected by printing to the output window.

So the problem I am having is when I call a 20 period ATR on my data-series I dont get the result expected. It is approximately half what it should be,

When I compair ATR(Closes[0], 20)[0] & ATR(hClose, 20)[0] here is a sample of what I get with the inputs to them:-



***********
***********
Current 1 Tick Bar 537582 ATR last 20 Closes 164.32 164.31 164.25 164.21 164.21 164.23 164.2 164.19 164.18 164.2 164.2 164.23 164.22 164.25 164.29 164.35 164.28 164.31 164.3 164.31
Current 1 Tick Bar 537582 ATR last 20 hClose 164.31 164.31 164.25 164.21 164.21 164.23 164.2 164.19 164.18 164.2 164.2 164.23 164.22 164.25 164.29 164.35 164.28 164.31 164.3 164.31

ATR 0.0730076458405191
ATR hCLose 0.0365543132310812
***********
***********
Current 1 Tick Bar 537583 ATR last 20 Closes 164.32 164.31 164.25 164.21 164.21 164.23 164.2 164.19 164.18 164.2 164.2 164.23 164.22 164.25 164.29 164.35 164.28 164.31 164.3 164.31
Current 1 Tick Bar 537583 ATR last 20 hClose 164.32 164.31 164.25 164.21 164.21 164.23 164.2 164.19 164.18 164.2 164.2 164.23 164.22 164.25 164.29 164.35 164.28 164.31 164.3 164.31

ATR 0.0730076458405191
ATR hCLose 0.0365543132310812
***********
***********

I cant figure out why they would be different ?

[koganam]

I have created a dataseries (hCloses) in my multi Series strategy ( 10min (primary) and 1 Tick(secondary)). The dataSeries is the same as Cloeses[0] apart from the Value[0] of the data series that is set to Closes[1]:-

if(BarsInProgress == 0)
				{
					hClose.Set(1,Close[1]);
				}
				
				if(BarsInProgress == 1 && CurrentBars[baseBars] > 1)
				{
					hClose.Set(Closes[1][0]);
                                }

I have verified the that both series are as expected by printing to the output window.

So the problem I am having is when I call a 20 period ATR on my data-series I dont get the result expected. It is approximately half what it should be,

When I compair ATR(Closes[0], 20)[0] & ATR(hClose, 20)[0] here is a sample of what I get with the inputs to them:-



***********
***********
Current 1 Tick Bar 537582 ATR last 20 Closes 164.32 164.31 164.25 164.21 164.21 164.23 164.2 164.19 164.18 164.2 164.2 164.23 164.22 164.25 164.29 164.35 164.28 164.31 164.3 164.31
Current 1 Tick Bar 537582 ATR last 20 hClose 164.31 164.31 164.25 164.21 164.21 164.23 164.2 164.19 164.18 164.2 164.2 164.23 164.22 164.25 164.29 164.35 164.28 164.31 164.3 164.31

ATR 0.0730076458405191
ATR hCLose 0.0365543132310812
***********
***********
Current 1 Tick Bar 537583 ATR last 20 Closes 164.32 164.31 164.25 164.21 164.21 164.23 164.2 164.19 164.18 164.2 164.2 164.23 164.22 164.25 164.29 164.35 164.28 164.31 164.3 164.31
Current 1 Tick Bar 537583 ATR last 20 hClose 164.32 164.31 164.25 164.21 164.21 164.23 164.2 164.19 164.18 164.2 164.2 164.23 164.22 164.25 164.29 164.35 164.28 164.31 164.3 164.31

ATR 0.0730076458405191
ATR hCLose 0.0365543132310812
***********
***********

I cant figure out why they would be different ?

You are resetting your hClose, so hClose[0] != Closes[0]. Ergo, things calculated on those series and indexed at zero will usually be different values.

[GKonheiser]

Hi koganam,

not sure I follow. I am resetting the value at index 0 when there is an 1 tick series changes but the it is then fixed to the close of the 10min series once it has closed.

You can see above the outputs from the two series. They are identical apart from index 1 , as I want, but it doesnt explain the massive difference in the 2 ATRs.

But ive been thinking about it and obviously the ATR uses Highs and Lows for Calculation. If I am basing my ATR on my series , hClose, what high and low is it using?

[koganam]

Hi koganam,

not sure I follow. I am resetting the value at index 0 when there is an 1 tick series changes but the it is then fixed to the close of the 10min series once it has closed.

You can see above the outputs from the two series. They are identical apart from index 1 , as I want, but it doesnt explain the massive difference in the 2 ATRs.

But ive been thinking about it and obviously the ATR uses Highs and Lows for Calculation. If I am basing my ATR on my series , hClose, what high and low is it using?

If you use a DataSeries as an input to an indicator, all Input is the value of the DataSeries. Therefore, in this particular case, as the NT ATR uses Wilder's Average, you are in effect calculating the Wilder's Average of the Close.

[GKonheiser]

So if I understand you correctly then when Closes[0] ( just for example) is passed to an indictor that needs also the high and the low, then it uses Hights[0] and Lows[0], is that what you mean ?

I thought that Highs[0] would be one data series and Lows[0] would be another?

So when you pass Closes[0] to the ATR say, it is calculating the atr based on the highs and the lows of that data series right ?

If that is the case how are Closes[0] and Highs[0] and Lows[0] etc.. linked ?

[koganam]

So if I understand you correctly then when Closes[0] ( just for example) is passed to an indictor that needs also the high and the low, then it uses Hights[0] and Lows[0], is that what you mean ?

I thought that Highs[0] would be one data series and Lows[0] would be another?

So when you pass Closes[0] to the ATR say, it is calculating the atr based on the highs and the lows of that data series right ?

If that is the case how are Closes[0] and Highs[0] and Lows[0] etc.. linked ?

No. A BarsArray is different from a DataSeries. A DataSeries has only one value per bar, so no matter what you use as Input in an indicator, if you pass the indicator a DataSeries for input, then the Input used by the indicator will identically be the value of the DataSeries. So regardless what references the indicator uses: Open, High, Close or Low, the indicator will use the value of the DataSeries for all of them. After all, the DataSeries, does not have any OHCL data, just a single value at any data point.

[GKonheiser]

OK got it. So the best thing to do is write my own method for this I think. Thanks very much for your help , I have leaned something new.

Is it possible to create your own Bars Array?

[koganam]

OK got it. So the best thing to do is write my own method for this I think. Thanks very much for your help , I have leaned something new.

Is it possible to create your own Bars Array?

Not in NT7. At least not without some serious undocumented stuff. And you should not need to do so for what you are doing. You have a BarsArray simply by declaring one. Your current code has 2 of them: the one associated with the time frame for the chart, and the one that you declared,

For processing purposes you could always declare any number of DataSeries to hold what you want to process and go from there. After all, a BarSeries does amalgamate 4 DataSeries into one object. You can construct your own object any way that you want to do.

[GKonheiser]

Yes exactly there are many ways to do it. Thanks again for the help, I understand now completely why it wont work as you have outlined the difference between a BarsArray and a Dataseries.

[NinjaTrader_PatrickH]

Thank you for helping out GKonheiser with their questions, koganam.

For any one else reading through this thread, koganam is correct in pointing out that the DataSeries used would not include High or Low values as it was not a BarsArray and a BarsArray could not be created in the supported realm of NinjaScript.