You can use both to generate indicator with different technique, in NS2 you may use prediction to a reference indicator, while in CH you can use either Buy/Sell cut off or indicator prediction.
Build Neural Network Indicator in MT4 using Neuroshell
You can use both to generate indicator with different technique, in NS2 you may use prediction to a reference indicator, while in CH you can use either Buy/Sell cut off or indicator prediction.
Hi jiacopel,
What you mean with range bars? can you put a sample chart? I never use noxa cssa with range bars.
You may put my indicator AT120EUG in your chart, then put trading strategy as shown on screenshot, the strategy work wells in all periods.
You may compare the equity with your original prediction strategy.
Hi Arryex!
see picture, when I use your indi, it has an other performance... maybe because I use other date (actual) and other settings. (optimization ends earlier...btw...)
When I use your orignal settings, the perfmance is not that good. Maybe
I've done somthing wrong... ;-) I'll check this again...
Bye, AT
Attachments
Last edited:
Hi At,
Check the setting on post #1018, third picture showing the parameters. You are using different values, example I use AT120EUG(Close, 3.152786E-03) while you are using AT120EUG(Close, 4600). I use similar parameter AT120EUG i.e. 3.152786E-03.
I am using your data while optimizing it, so there should be no different result between yours and mine. To ensure it, just use my parameter setting and give NSDT a free optimizing for example on the threshold values.
To be noted, I use same threshold for Long Entry and Long Exit, Short entry and Short Exit, while you are using non similar values of thresholds. Hope it will works.
Check the setting on post #1018, third picture showing the parameters. You are using different values, example I use AT120EUG(Close, 3.152786E-03) while you are using AT120EUG(Close, 4600). I use similar parameter AT120EUG i.e. 3.152786E-03.
I am using your data while optimizing it, so there should be no different result between yours and mine. To ensure it, just use my parameter setting and give NSDT a free optimizing for example on the threshold values.
To be noted, I use same threshold for Long Entry and Long Exit, Short entry and Short Exit, while you are using non similar values of thresholds. Hope it will works.
Hi arryex!
Thanks for your feedback!
I tried (different) settings, and also full optimization. (sometimes "opti. ended up earlier...")
I think the different between the results occur because of different dates?!
Also with original settings - but actual date - I get different results, see picture... original.jpg. Some different settings:
Opti_1.jpg -> setting in search space: 0 - 50 (long) and -80 - 0 (sell)
Opti_2.jpg -> setting in search space: 0 - 50000 (long) and -80000 - 0 (sell)
Opti_3.jpg -> setting (fixed): 30000 - 30000 (long) and -1200 - -1200 (sell)
Date always the same makes no really different wether I use 1 week or 2 weeks...
Thanks for your feedback!
Bye, AT
Attachments
Hi At,
Please find here attached chart complete with trading strategy (within the zip file), you may check what are the reason your strategy is different with mine. Please select "Yes" when use saved data in chart question dialog shown.
To be noted, on this sample we are using 3 weeks data for optimization, 3 weeks on paper trade and 1 week for OOS. This strategy may required for retrain in case a new data arrive providing bad result.
Please find here attached chart complete with trading strategy (within the zip file), you may check what are the reason your strategy is different with mine. Please select "Yes" when use saved data in chart question dialog shown.
To be noted, on this sample we are using 3 weeks data for optimization, 3 weeks on paper trade and 1 week for OOS. This strategy may required for retrain in case a new data arrive providing bad result.
Hi Arryex!
Many many thanks for your help!
It works for me, same time/date -> same results.
As soon as I change the time/date -> bad results...
How do i make the retrain, as the indi you sent me, is a little bit a blackbox for me ;-)) Could you help me or tell me how you did it?
A idea to get better results is, to use the jma with different time settings. e.g. 9, 15, 20 or something like that.
Thanks a lot and bye, AT
Hi AT,
It is good if you already get the same result.
The trick that i used is by searching indicator and threshold values to predict future open (or percent change in Open). The simplest one is by using Chaos Hunter, but if you known the techniques you may use other software.
Actually, Neuroshell Trader prediction strategy is using the similar technique with more available options (both input and output), you can use future values of indicator as output prediction and let NST to find the best threshold.
It is good if you already get the same result.
The trick that i used is by searching indicator and threshold values to predict future open (or percent change in Open). The simplest one is by using Chaos Hunter, but if you known the techniques you may use other software.
Actually, Neuroshell Trader prediction strategy is using the similar technique with more available options (both input and output), you can use future values of indicator as output prediction and let NST to find the best threshold.
Hi Arryex!
I think I got the point with the threshold... I'll see. ;-)
Do you use CH for predicting the future open? (I think no)
I think you use CH for optimal threshold vales and then use NS for e.g. 5 bar predicting the future open%?
How have you made the AT120EU Indi?
Actually, I used more than one JMA setting (9,11,15,23,etc...) and let the optimizer select the inputs for trading... but with no good results... (testing...)
Thanks a lot and bye, AT
Hi Tovim/AT,
This chart showing an indicator that has been built from four JMANS indicator, with parameter 8,21,34, 55 and all phases set at 0.
Actually in CH, we can set the Open as output, unfortunately it is not clear what is the exactly the parameter their use (it may be optimal change, %change, etc).
To predict future %change Open is similar trick used for other software, you can put the future %change as output and let CH to find indicator prediction.
Other news, new Nutonian Eureqa is no more free, they asked for about $2300 for a license. I think this is quite expensive compare to CH.
This chart showing an indicator that has been built from four JMANS indicator, with parameter 8,21,34, 55 and all phases set at 0.
Actually in CH, we can set the Open as output, unfortunately it is not clear what is the exactly the parameter their use (it may be optimal change, %change, etc).
To predict future %change Open is similar trick used for other software, you can put the future %change as output and let CH to find indicator prediction.
Other news, new Nutonian Eureqa is no more free, they asked for about $2300 for a license. I think this is quite expensive compare to CH.
Attachments
Hi Arryex!
Looks nice so far ;-)
Actually I have tested the indi but no idea, how you manage to make one indi.
like ou did with the threshold. Also no idea how the get the TS with CH. ;-)
I'm testing and trying different things (MACD, CCI, JMAN, etc...) with CH...
Bye, AT
Hi AT,
I think all the user has did a similar mistake, we want to get immediate profitable trade without knowing how to work with the software properly.
My short advice (but it should be powerful ha..ha...), since you own CH, please open help file, Chapter 6 Examples, Example 5 - Running ChaosHunter Models in NeuroShell Trader
This should clearly explain what you should do to get CH modelling and display it on NSDT.
You can put hd.csv in your NSDT data folder, then follow the step.
Hope you will be successful.
I think all the user has did a similar mistake, we want to get immediate profitable trade without knowing how to work with the software properly.
My short advice (but it should be powerful ha..ha...), since you own CH, please open help file, Chapter 6 Examples, Example 5 - Running ChaosHunter Models in NeuroShell Trader
This should clearly explain what you should do to get CH modelling and display it on NSDT.
You can put hd.csv in your NSDT data folder, then follow the step.
Hope you will be successful.
Hi Arryex!
Read Chapter6 -> example5 is exactly what I'm going to do!
+ also testing and try to understand it! ;-)
CH -> MT4 -> formular integrated as DLL? or with the "CH/MT4 bridge"?
For the next days I'm going to test and try some "setups"...
Thanks for help and infos!
Bye, AT
BTW: How did you produce this AT120EU indi? ;-)
Hi AT,
You can download the code to integrate CH into MT4 (MT4 add on) either to show output indicator, trade signal or equity with two ways dll (CH run time) or inserting formula from ward.net.
There is limitation using dll (see CHpredict sample) or CHruntimelibrary, you can only insert the indicator with OHLCV raw data basis. But so far I can create any CH formula even with MT4 indicator as input and not limited to OHLCV data only, this is the technique I use.
To insert AT120EU coming from CH, I convert directly the CH formula into NSDT indicator but you can also using external dll indicator using CH model, if you want.
You can download the code to integrate CH into MT4 (MT4 add on) either to show output indicator, trade signal or equity with two ways dll (CH run time) or inserting formula from ward.net.
There is limitation using dll (see CHpredict sample) or CHruntimelibrary, you can only insert the indicator with OHLCV raw data basis. But so far I can create any CH formula even with MT4 indicator as input and not limited to OHLCV data only, this is the technique I use.
To insert AT120EU coming from CH, I convert directly the CH formula into NSDT indicator but you can also using external dll indicator using CH model, if you want.
Hi Arryex!
Alright I have to check this. A lot of new stuff (DLL, etc...) for me.
I think I'm going to either convert fomular to NSDT or use the import function of NSDT, CH signal...
At the moment I'm trying to find an profitable model in EU 1H with CH, with
CCI, ADX, and JMAN but my problems is -> either bad model... or a PC which is tooooo slow... ;-)
how have you made your AAJMAN indi, only different times (13,21, 34, 55,) of JMAN + crossover strategy?
Thanks a lot for help and info!
Bye, AT
Alright I have to check this. A lot of new stuff (DLL, etc...) for me.
I think I'm going to either convert fomular to NSDT or use the import function of NSDT, CH signal...
At the moment I'm trying to find an profitable model in EU 1H with CH, with
CCI, ADX, and JMAN but my problems is -> either bad model... or a PC which is tooooo slow... ;-)
how have you made your AAJMAN indi, only different times (13,21, 34, 55,) of JMAN + crossover strategy?
Thanks a lot for help and info!
Bye, AT
Hi Arryex!
As you see I have my "problems" in finding a good model... ;-))
Inputs: CCI, ADX, JMANS55, JMANS7, 2000 rows H1 EU (backtesting)
Actualy CH only use, CCI and ADX as input...
Output: Open (OOS: 300 rows EU H1)
Any ideas, how to say CH, it have to use parameter/input X?
Bye, AT
As you see I have my "problems" in finding a good model... ;-))
Inputs: CCI, ADX, JMANS55, JMANS7, 2000 rows H1 EU (backtesting)
Actualy CH only use, CCI and ADX as input...
Output: Open (OOS: 300 rows EU H1)
Any ideas, how to say CH, it have to use parameter/input X?
Bye, AT
Attachments
In my opininon that means your inputs( CCI, ADX, JMANS55, JMANS7) are not sufficient enough to create a profitable model.CH only use CCI and ADX as input because they are the most profitable inputs.
The thing is CH only creates an indicator out of your inputs.So it doesnt create a full model.I have tried other genetic programming softwares where you have monte carlo simulation , money managment rules,risk managment rules etc.every software have their own problems.
The thing is CH only creates an indicator out of your inputs.So it doesnt create a full model.I have tried other genetic programming softwares where you have monte carlo simulation , money managment rules,risk managment rules etc.every software have their own problems.
