EMA ERRORS

Section for discussions related to indicators, use of indicators, and building of trading stategies using indicators.

Moderator: admin

EMA ERRORS

Postby Bebbspoke » Sat Aug 27, 2011 10:11 pm

Please could the Marketscope display of INCORRECT emas be REMOVED or CORRECTED?
I am fed up with reporting the howling ERRORS in that WRONG values are shown on the lefthand side of the chart which give totally MISLEADING information.
it appears that the software writers are unable to rectify the fault - how many other mathematical ERRORS lie hidden in the package??? - I CANNOT trust such a display.
The Forex market is by nature fickle and unpredictable - surely there is no need for Marketscope to make things WORSE? (please do not argue that statement - if it were not true then all traders would be billionaires...)
User avatar
Bebbspoke
FXCodeBase: Confirmed User
 
Posts: 170
Joined: Tue Aug 16, 2011 4:04 am
Location: Walton on the Naze UK

Re: EMA ERRORS

Postby sunshine » Sun Aug 28, 2011 8:02 am

We greatly appreciate feedback and error reports and will look into all reported issues.

Could you please kindly explain which exactly issue you experience with Marketscope.
Do I understand correctly that you mean that EMA indicator (included in the standard set of Marketscope indicators) is calculated incorrectly? If so, could you please kindly show the example (with exact values which are incorrect and how they should be).

Just in case here is the EMA formula used in Marketscope
http://fxcodebase.com/wiki/index.php/Ex ... erage_(EMA)
By the way, Marketscope EMA is calculated in the same way as EMA in other popular charting applications (for example, Meta Trader).
sunshine
 

Re: EMA ERRORS

Postby Nikolay.Gekht » Tue Aug 30, 2011 11:20 am

We discussed this problem in email, but, probably, the result would be interesting for others as well.

The problem is that EMA depends on a big number of historical data, definitely bigger than N. for example, if you apply EMA(10) on 5000 bar history from the beginning and from the middle, the result will be different for the first approx 20 in the beginning of the second indicator, so these data are not reliable. In light of the fact that current default size of the history is 300 bars and EMA is often used with N=100+, such EMA data may be completely unreliable. BTW, the same is related for MACD and other exponential-based indicators.

I am discussing the solution, proposed by John Bebb in email - preloading of enough amount of data to make such indicators reliable from the start, with developers right now. I'll do my best to include this solution into the next release or over the release in the worst case.
Nikolay.Gekht
FXCodeBase: Site Admin
 
Posts: 1235
Joined: Wed Dec 16, 2009 6:39 pm
Location: Cary, NC

Re: EMA ERRORS

Postby dtb71fxcm » Sat Nov 23, 2013 8:22 pm

Has there been any work on this?
This is critically important -- especially for strategies whom incorporate the MAs like EMA
User avatar
dtb71fxcm
 
Posts: 44
Joined: Sat May 05, 2012 9:28 pm

Re: EMA ERRORS

Postby Bebbspoke » Sun Nov 24, 2013 10:34 am

Hi there - I believe work is in hand - investigations have disclosed additional areas for possible distortions of the EMA graphs; - such as the use or "non-use" of the "Weekend//Sunday//Silent" times and/or values. I am researching into historic data trending - some of the "historic data" processes are under scrutiny and known to be problematic.
Cheers, Bebbspoke
User avatar
Bebbspoke
FXCodeBase: Confirmed User
 
Posts: 170
Joined: Tue Aug 16, 2011 4:04 am
Location: Walton on the Naze UK

Re: EMA ERRORS

Postby Valeria » Tue Nov 26, 2013 4:22 am

Hi dtb71fxcm,

Has there been any work on this?
This is critically important -- especially for strategies whom incorporate the MAs like EMA


It is implemented for the charts. When a strategy is started, minimum 300 ticks are available. But if the strategy downloads the historical data itself, it should download enough data to avoid the described problem.
Valeria
 

Re: EMA ERRORS

Postby Bebbspoke » Mon Apr 21, 2014 7:52 am

In 2011 I observed errors in the (Market Scope) EMA generation and subsequent to email discussion,
I was contributory to the developers considerably improving the situation.
I believe a pre-load of 300 pips is now used, however - this correction is still not perfect: -
The pre-load quantity is still inadequate with regard to the 200EMA, and may still produce small errors in the 50EMA - because we are actually dealing with six digit numbers, in a 200EMA even a pre-load of 700 bars may still procure error at the 0.1 pip level.
It is unlikely that errors of a few pips with indicators as sluggish as the 200EMA or 50EMA may be considered as very relevant, except perhaps in the determination of specific cross / uncross events,
- but for mathematical correctness I do now have a technique which much reduces the pre-load requirement, should anyone be interested.
Whilst on the subject of correct mathematical treatment of indicators I have since observed that the Market Scope derivation of non-default ATR"n" values appears erroneous - I have started a new thread as appropriate.
User avatar
Bebbspoke
FXCodeBase: Confirmed User
 
Posts: 170
Joined: Tue Aug 16, 2011 4:04 am
Location: Walton on the Naze UK

Re: EMA ERRORS

Postby Valeria » Mon Apr 28, 2014 2:24 am

Hi Bebbspoke,

Thank you for the suggestion. We are currently considering it. I or my colleagues will get back to you when we discuss it.
Valeria
 


Return to Discussions

Who is online

Users browsing this forum: No registered users and 3 guests