Ok, this is a cross post..
But I think this might be valuable for everyone's understanding...
Originally from the Great HGB forward test topic...
This all to try and find an explanation why the results differ with the forward test, of the same EA, Same Setfile, and same broker.
Just want to share some interesting stuff.
I've also experienced that not all signals are taken.
This caught my attention and curiosity.
In the HGI indicator topic was some conversation about repainting.
And the change of signals, don't want to dig to deep into it here.
But this might be an interesting read for you.
There was also someone (Drunka) http://www.stevehopwoodforex.com/phpBB3 ... 80#p112158
who had multiple instances of Empty4 and the indicator.
On the same broker, but with different signals on the charts.
As we don't know, how the HGI calculates stuff, maybe this is caused by some interpolation or extrapolation of values??
Either way interesting....
Yes you are really on to something with the HGI Readdelay.WhoKnows » Tue Jan 20, 2015 10:42 pm wrote:
The second point would be the HgiReadDelaySeconds setting. The larger this setting, the more chance there is that a signal that comes and disappears within the ReadDelay period is missed. And with different computers looking at different timestamps within the ReadDelay period... different computers may miss different temporary signals.
It would be interesting to see what effect increasing the HgbReadDelay has on the performance of HGB. Maybe take it from 60 too 300 seconds to weed out bad signals .... On a 4H trading TF 5minutes is not much. Need to think this through. Just an hypothesis, no proof.
Im experimenting with this. Did set it, so that at least 10 checks per candle.
But when trading H4, this still might be to long distance apart.
The interpretation of the signals is one of the reasons why I switched with the BoHGB from using the indi to the library.
Oh, just have an insight..
The next indi read time is determined from the moment the EA is initialized.
It is very very unlikely that you all attached the EA at the same time to the chart.
And clicked simultaneously on the OK butting, for the settings.
So, all of your indicator read-times are different from start, and stay different from that moment on.
Depending on the settings it could be seconds of, or even milliseconds.
Coding the EA a bit different, MIGHT improve this, but I don't believe it possible to get this thingy entirely out of our way.
Today I did some eyeballing on the charts...
I saw a sell wavy drawn on the chart at the current candle, and all of a sudden it was changed to a buy wavy. You clearly didn't want to enter that one as a sell.
The sum this up.. and combine this with my previous statement.
The way your instance of HGB took the trade is entirely depending on the time the signal from that candle is read. And as stated above, this differs from each and everyone of you, and each instance.
I also noticed, sometimes trades where closed by an opposite signal.
Even when there was no visible opposite signal on the chart.
(Have to add, that this could also be the case because I use the LIB and indicator in the mix.)
Just want to make clear, this is no bug at all! This is normal behavior.
It is the result of how the HGI and HGB works.
A discretionary trader won't notice even notice this change of signals, or a bad signal for a short time.
But an EA reacts much faster in the milliseconds area.
This is one of the challenges someone faces when automating stuff..
It will be hard to overcome this, but at least there can be something done to try to improve this.
It is common practice to use filters for this purpose.
I have a background in programming physical stuff.. over there it's common practice to use this.
Dunno if its common for MQL programmers.
At this moment the HGB and my version the BoHGB doesn't have this kind of filter in place.
This is one of the things I wanna pay attention to and work on.
Before I take the next steps..(different strategies ect..)
Will have to add that this is THE challenge.
Discretionary trades have such an easy job, not to deal with this kind of stuff
.Because the text above was only focused on the possibility of a "wrong" signal or changing signal appearing on the candle. Have to add, that there is also a the repainting thingy.
Won't go deep into this one, but as reported on the HGI indicator thread.
Sometimes a signal is formed on the previous candle. (on H4 this might sometimes be to late to enter when this happens).
Well, all this said, we can have multiple approaches with this.
Accept that the HGI has it's weaknesses. (if I may call it so).
And the same goes for the HGB, BoHGB.
Accept it, leave the system simple and enjoy it...
You'll never gonna get a 100% perfect system...
Personally have to say I am the kind of person who likes a challenge and is intrigued by this.
Was already on the topic today in some PM's so good to share this here.
I accept the challenge to see, if I can come up with some form of filter to (potentially) reduce these kind of unwanted system behavior.
Will create a separate EA, with only purpose to catch this type of behavior, to see what we have to work with.
Edit: have to add that this might be over optimization.. we'll see.
Hopefully your are not bored with all this tech talk
