Stochastic over indicator

[NinjaTrader_JessicaP]

I am happy to clarify manugarc and I apologize for the confusion.

There were two things occurring in this thread :

• K and D values out of range were occurring in the stochastics indicator
• Input values other than price values were being ignored

These two are related. By fixing the first item, we progressed and ran into the second item. Harry had been working with some of my colleagues on the second item, and was given a tracking ID. I placed a feature request very similar to that feature request, so that we could address the second item.


Please use the thread http://www.ninjatrader.com/support/f...ad.php?t=90105 for further discussion regarding the second item.


Please use this thread should you receive a k or d value below 0 or above 100 again.

[manugarc]

Hi Jessica,

I'm sorry, I'm so confused.

From post #21:

I apologize for the lack of clarity. To clarify, I have access to newer, in-development versions of NinjaTrader, and when I was testing using one of these versions, the behavior you and I observed stopped occurring in these versions. When the next version of NinjaTrader is released, what you observed with regard to K and D out-of-range values should be fixed. If this is not the case, please let us know so we can track down why.

We were talking about version 8.0.5.0, so I was quite sure that version 8.0.5.0 would have this problem fixed, and it hasn't (if I'm not doing something wrong, but I've recompiled everything).

From your last post:

This customer would like a user to be able to select the mapping for each input value to an indicator

Thats not correct; I do am able to select the input of the indicator, the problem is the indicator ignores it from version 8.0.5.0. On previous versions there was not this problem.

So, I don't understand it. If you were testing it on 8.0.5.0 and everything was OK, why it's not working for me now that I am using this version? That's why I asked whether I was doing something wrong.

[NinjaTrader_JessicaP]

I will need to mention at this time that the product management team will be making the decision as to whether to include user code.

If you have no further information regarding differences between 8.0.4.0 and 8.0.5.1, or k or d values leaving the 0-100 range, I will close this ticket.

[Harry]

Just checked, I have already recoded

- ADX
- ADXR
- ATR
- CCI
- DM
- Stochastics

to work with an indicator as input series.

No problem to discuss the issue with someone who wants to do the remaining ones.

[NinjaTrader_JessicaP]

Thank you Harry.

To clarify my wording, I intentionally picked wording that was similar to SFT-1606 so that our product management team would be able to locate this ticket easily. This will make it easy to track things on our end, and will ensure the fastest possible service.

In addition, since I have put in a feature request, and have closed my investigation into NTEIGHT-11194 (k and d values leaving range), so that all of our tracking tags now refer to SFT-1606, all information you add to this thread will be made immediately available to the product management team, including your most recent post.

[Harry]

Hi Jessica,

The misunderstandings continue.

This customer would like a user to be able to select the mapping for each input value to an indicator

I definitely do not want any changes to the mappings. This is not necessary. The mapping is perfect the way it is. Don't change it, it may only get worse.

What I am asking for is:

Please recode your indicators to work correctly!

Recoding of the affected 10 or 20 indicators can be easily done within a few hours. Please do not complicate things, I am not talking about any changes in the architecture. I am just asking for inserting a few lines into the code of each indicator, just to make them work correctly.

I am even ready to recode them for you.

This is a quick hit, easy to understand, easy to do. No efforts!

[NinjaTrader_JessicaP]

Thank you Harry, I have made myself aware of the situation with SFT-1606, and your video is very much appreciated and makes the situation immediately clear. I have included this video in with our notes so that future techs assisting you will be able to provide you with faster assistance.

The misunderstanding was caused by your mention that 8.0.4.0 and 8.0.5.1 work similarly. This should definitely not be the case. A bug unrelated to what you are discussing, which was allowing for out-of-range k and d values in some circumstances, was uncovered and repaired between these versions.

You can see post 16 for more information regarding NTEIGHT-11194 , which is unrelated to SFT-1606

manugarc, pertaining to the feature Harry mentioned, I have submitted a feature request to the product management team for the following feature :

This customer would like a user to be able to select the mapping for each input value to an indicator

I will follow up with more information as soon as it's available. If the feature requests already exists, a vote will be added to it.


Please let us know if there are any other ways we can help.

[Harry]

Jessica,

First of all thank you for your quick answer. Sometimes it is really painful to write here, because it takes ages before things are moving ahead ...

Let me walk you through this here:

(1) Unlike NinjaTrader 7, NinjaTrader 8 does not remap Open/High/Low/Close/Median/Typical to Input. This is a change in design, and it is a change in design for the better. It is an advantage because you can now easily access both an indicator series and a price series without adding anything else!

(2) The NinjaTrader Stochastics and many other NinjaTrader indicators (see list posted below) have not been adapted to the change in design and cannot be used with input series other than price.

(3) It is easy to recode them. Actually I have already recoded most of them to make them usable with other input series. I have also posted a version of the Stochastics that works as expected.

As you ask for a video, I have created a video here to show you that the Stochastics does not work as expected.

https://www.screencast.com/t/5nInH2h89c

Please let me know, whether this clarifies the problem.

[NinjaTrader_JessicaP]

I have explained in post #23 of this thread that the Stochastics and other indicators cannot be used with any input series other than price. These indicators need to be recoded. This has nothing to do with NT 8.0.5.0.

This is not what we observed in our testing on our end.

So that I may have a more complete picture of what is happening on your end, would it be possible to use the freely and publicly available screen capture software Jing, https://www.techsmith.com/jing.html , or a similar program which can record video of what is happening on your screen, so that you can show me what is occurring on your end? These quick instructions can get you started with Jing.

1. When you start Jing, a yellow half circle will appear near the top of your screen. Please hover over it with your mouse
2. A small + sign will extend out of this half circle. Please click on it
3. When your cursor becomes two intersecting lines, please click and drag your mouse over an area of your screen you would like to record
4. Please press the film strip button that appears to begin recording
5. When you are finished recording, please press the square stop button that appears
6. Please press the share button that appears. It consists of three vertical upward pointing arrows.

You will then have the option to view your video on screencast.com . This is the link which can help me diagnose what is happening on your end. It is possible you will need to set up a free account in order to share videos.




Thank you very much in advance for providing us with this information. If this procedure is not an option for you, or if there are any other questions we can answer, please let us know.

[Harry]

Hi NinjaTrader_JessicaP,

I'm trying version 8.0.5.0 and now Stochastics always use price as input, ignoring the inputs I set (as other indicators instead of the price itself), Am I doing something wrong?

Thanks in advance.

I have explained in post #23 of this thread that the Stochastics and other indicators cannot be used with any input series other than price. These indicators need to be recoded. This has nothing to do with NT 8.0.5.0.

- ADX
- ADXR
- Aroon
- Aroon Oscillator
- ATR
- CCI
- DM
- DMI
- Double Stochastics
- Parabolic SAR
- RVI
- Stochastics
- Stochastics Fast
- Ultimate Oscillator
- WilliamsR

Here is a link to a modified Stochastics.

http://ninjatrader.com/support/forum...81&postcount=6

[manugarc]

Hi NinjaTrader_JessicaP,

I'm trying version 8.0.5.0 and now Stochastics always use price as input, ignoring the inputs I set (as other indicators instead of the price itself), Am I doing something wrong?

Thanks in advance.

[manugarc]

Hi Harry; I'm sorry I haven't had time to check that; you may be right, but that code would highly overload the processor, using trys and catchs on each tick, beside the multiple checkings. Thank you for the idea anyway.

[Harry]

Hello manugarc,

If you look at the code of the Stochastics, it is definitely impossible that fastK returns a value outside the boundaries [0,100]. I think we are safe to assume that fastK works correctly.

This means that the problem arrises when fastK is being smoothed with the SMA. The SMA uses a recursive formula. The recursive formula calculates much faster than the standard formula (adding up the last N values and dividing them by N), but is more sensitive to error propagation, as rounding errors are adding up.

As we take into account the SMA formula, there are basically two possible source for an error.

(1) If you use the Stochastics with a period larger than 256, the value of fastK is not defined 256 bars ago, unless you set all Series to MaximumBarLookBack.Infinite.

In order to avoid that problem, you should change the code to:

den = new Series (this, MaximumBarsLookBack.Infinite);
nom = new Series (this, MaximumBarsLookBack.Infinite);
fastK = new Series (this, MaximumBarsLookBack.Infinite);

(2) In case that you use small bar periods, for example 30-tick bars or similar, the problem of the shift maybe caused by the error propagation properties of the recursive SMA. In this case I would suggest to replace the SMA with a conventional SMA that calculates the current value by adding up the last N values of fastK and dividing them by N.

Again, solution (1) is applicable in case that you used a period larger than 256. Solution (2) is applicable in case that you used the Stochastics on a high resolution chart over a longer time period.

Can you reproduce the problem?

[manugarc]

Hello Harry,

Actually I used a modified code, replacing OHLC values with "Input" (as I pointed on my first post) and it failed too. I can't say your code will also fail, but I'm pretty sure it won't fix the problem.

[NinjaTrader_JessicaP]

Thank you for this additional information Harry. My colleague NinjaTrader_ChelseaB is currently out of the office. I will review SFT-1606 with him upon return, and ensure that the behavior we discussed with respect to stochastics producing out-of-range values remains corrected while SFT-1606 is implemented.