Empirical Mode Decomposition update(?)

[Trader.Jon]

NinjaTrader_Josh,

I have been testing on the Empirical Mode Decomposition you have kindly provided for us in the downloads section. I like what I see, and have a question on use for strategy building.

On occasion during backtesting, (about 2% of the time), going short will cause an 'object reference' error, but for longs this never happens. I isolated the lines to when the code is verifying that the MainPlot is falling ie

(MainPlot[0] < MainPlot[1] && MainPlot[0]< LowerBand[0])

removing that code or replacing it with

(MainPlot[0] > MainPlot[1] && MainPlot[0] > LowerBand[0])

stops the errors, so I am (very much only) speculating that it has something to do with the floating point math and negative values (not my area of expertise). My request is some code update to allow a normalization of the values for the indicator output, such that they are all positive. My speculation is that would resolve the issue.

Of course, other improvements suggestions or fixes are appreciated. I am using NT7rc1, and the indicator was built in NT6.5.

Thanks in advance,
Jon

[NinjaTrader_Bertrand]

Jon, I'll check with Josh and let you know our thoughts.

Thanks,

[Trader.Jon]

Bertrand,

Thanks for the assist.

Jon

[NinjaTrader_Bertrand]

Jon, can you please check if you run into the same issues on looback set to Infinite?

[Trader.Jon]

Bertrand,

Makes no difference.

Jon

Jon, can you please check if you run into the same issues on looback set to Infinite?

[NinjaTrader_Bertrand]

Thank for letting me know, the same outcome if you hardcode the infinite setting needed into the indicators Initialize()?

http://www.ninjatrader.com/support/helpGuides/nt7/dataseries_class.htm

[Trader.Jon]

Thanks for the 'infinite' suggestion ... Makes no difference ... still a problem exclusively with short side: I removed all but the 'essential' code from the strategy and it runs and no problems with LONG trades, but as soon as I uncomment the code for the SHORT trade entry rules I get those object reference errors.

I find it odd that that kind of an error would actually persist only on SHORT, that is why I am hedgeing that it is do do with the negative#s ... maybe someday Ill try Math.Abs() on the short calculations and see if that makes a difference.

Jon

[NinjaTrader_Brett]

Hello,

I will have bertrand or Josh respond to this when he gets back into the office tomarrow since he has been working with you on this ticket.

Let me know if I can be of further assistance.

[NinjaTrader_JoshP]

Trader.Jon,

Can you please provide us with your test setup as I am unable to get the error message you mentioned? What instrument, dates, period type, etc. are you testing on? Thank you.

[Trader.Jon]

Hi Josh,

I appreciate the continued assist ... as a heads up I am still wanting to resolve the issue and I am putting it on a lower priority so responses will be affected by time available .. and I DO want to find a solution

I am testing on the whole month of January 2010, but the optimization can kick out in as little as one day (Jan 4) with the error, and wont proceed further. $EURUSD 3 range on the trading, but the EMD is running on BIP==1 (4 range).

TY!
Jon

[NinjaTrader_Bertrand]

Jon, thanks - we'll test this setup here and will look into shortly. Would you mind sending in the script you used to us to support at ninjatrader dot com?

[NinjaTrader_JoshP]

Trader.Jon,

We have analyzed the Empirical Mode Decomposition indicator itself and found no issues. As Bertrand mentioned, we will need your exact strategy utilizing the indicator to be able to assist further. Thank you.

[Trader.Jon]

Josh & Bertrand,

I have completely written a strategy from scratch, and the EMD (including my modified version) now is stable without errors. I do not have a clue where the error came from, and thankyou for your time and patience. Sorry for taking up your bandwidth. Will let you know if something suddenly appears ...

Jon