"Phantom" Close!! | Forex Factory

Options

Permalinks
Print Thread

Similar Threads

Phantom 6 System 1,540 replies

Phantom 6 - Manual System 8 replies

Phantom Entry on ODL Japan. 2 replies

Tradeview CAD/JPY "Phantom" spike 0 replies

Phantom high, phantom low 0 replies

"Phantom" Close!!

  • #1
  • First Post: Nov 6, 2009 3:19pm Nov 6, 2009 3:19pm
Hello all,

I need to resolve a problem that is difficult to describe so bear with me a moment. I wrote an EA that enters a position based on a close beyond a predetermined threshold. The EA monitors the close iClose (NULL,0,1); for an entry trigger beyond the threshold.

Every so often a "Phantom" close price will trigger an entry. I call it "Phantom" because the close price that triggered the entry is NOT the close price on the chart OR the close price that shows on comments that monitor that. It is abolustely wrong. I programed the EA to print the closing price and threshold upon entry and have found that when this happens the closing price triggering the entry was completely wrong.

When I try to reproduce the problem, (by taking the EA off the chart and reapplying it), nothing happens the EA works as it was programmed. So I cannot diagnose what is causing the problem. VERY FRUSTRATING.

The bad part is when this happens on a live chart I am stuck closing out a loser (messing up my win stats) when the logic should not have allowed an entry in the first place were it not for this "Phantom" number.

So my question is...where does this number come from? Does the data stream or a chart sometimes generate a random number? Have any of you encountered anything like this? Could there be a small corrupt character in the EX4 file when I compile?

Can anyone suggest any ideas?
  • #2
  • Edited 6:11pm Nov 6, 2009 4:53pm | Edited 6:11pm
An obvious question to me is why use iClose instead of Close[1]? I doubt this will solve the problem but I had to point it out because it is not standard coding practise.

Another suggestion, if these values are so random and unpredictable, why not put in some protection code like checking the value against a moving average of close prices? If the value is out of range just ignore it.
  • #3
  • Edited 6:23pm Nov 6, 2009 6:09pm | Edited 6:23pm
Quoting gspajon
Can anyone suggest any ideas?
I assume you want the close of a bar. In MT4 this is the last tick that would belong to this bar and unfortunately it is not easy to trigger something on the last tick because you never know if it really was the last tick that just came in, instead you could slightly modify your strategy and do something like the following to capture the open of the next bar instead.

The following is part of the skeleton i use for every new expert advisor or indicator i write:

Inserted Code
[color=SeaGreen]/**
* will be called on every tick
*/[/color]
void onTick(){
  [color=Olive][color=SeaGreen]// do something here on every tick
  // Bid and Ask contain the current prices for this tick[/color]
[/color]}

[color=SeaGreen]/**
* will be called on every open
*/[/color]
void onOpen(){
[color=SeaGreen]  // do something here on every bar open
  // Bid and Ask contain the current bar's open prices
  // you can use Close[1] to check the exact value of last bar's close
  // but remember: its already one tick later[/color]
}

[color=SeaGreen]/**
* start() is only used to call the onTick() and onOpen() functions.
* All action should happen in these functions.
*/[/color]
int start(){
   static int numbars;
   onTick();
   if (Bars == numbars){
      return(0);
   }
   numbars = Bars;
   onOpen();
   return(0);
}
  • #4
  • Nov 6, 2009 10:30pm Nov 6, 2009 10:30pm
Interesting answers...thank you both. CodeMeister, I use iClose because the time frame is variable depending on the volume of the monitored chart. It dynamically switches the time frame, which up until now, has never caused a problem.

On the last occurrance, i found that the closing price in question did exist in the history. But it was the closing price from 9 hour before. And now I am really scratching my head. How is it possible that that particular price (which by the way was over 20 pips away from the market at the time) showed up as a current close?

None of these "Phantom" prices are even close to where the market transitioned from one time frame to another. If it was the difference of even the spread I could deal with that...like 7 bit pointed out... but this has me pulling out my hair, what little is left.

I have temporarily programmed a comment screen with a bool to by pass on return (0). that displays the important numbers as read by the EA before allowing any further live trading. This helps a little but I would still like to know where these anomolies come from.

Is it possible that a time glitch in reading the history could cause the EA to mistakenly pick up the number from the wrong time frame, and then on the next tick find the right number? That one "bad pick up" caused the entry logic to trigger.

This is very elusive.
  • #5
  • Nov 7, 2009 7:20am Nov 7, 2009 7:20am
Quoting gspajon
On the last occurrance, i found that the closing price in question did exist in the history. But it was the closing price from 9 hour before. And now I am really scratching my head.
I suspect there is still one of these hard to find bugs in your own program where some variable is not updated correctly, initialized wrongly, one of these malicious typos (i instead of 1), something that happens only in a seldom used if branch, something that looks perfectly ok when staring at it for hours but is still wrong. I know these kind of bugs, you can waste days finding them, and after finding them they suddendly become so obvious that you don't understand how you could overlook it.
  • #6
  • Nov 7, 2009 10:27am Nov 7, 2009 10:27am
7-bit

Thanks for the insight...I'll look again...and again. I am just not seeing it. Admittedly I am not a programing expert but I've been doing mql now for quite some time. The program itself is quite simple really, so there is not much to go wrong.

I'll go through it character by character again...thanks
  • #7
  • Nov 7, 2009 11:12am Nov 7, 2009 11:12am
Quoting gspajon
I'll go through it character by character again...thanks
Make heavy usage of Print() for debugging. Print() is your best friend. When you suspect that iClose(foo, bar, shift) gives wrong data then print out before *every* call to the function

foo, bar, shift // to see what you are actually feeding to this function
iTime(foo, bar, shift) // to see what time this offset would mean
iTime(NULL, 0, 0) // to see the current time: when did this actually happen

Since you are dealing with different time frames, one thing comes to mind: the shift parameter is different for every timeframe. Use iTime() to get the absolute time corresponding to a shift in one timeframe and then use iBarShift() to find the correct shift for this time in another timeframe. You cannot just divide by 15 to translate M1 to M15 shifts. Always make the additional step via iTime() and iBarShift() and use absolute times to find corresponding bars in higher timeframes.
  • #8
  • Nov 8, 2009 5:16pm Nov 8, 2009 5:16pm
Quoting 7bit
I suspect there is still one of these hard to find bugs in your own program where some variable is not updated correctly, initialized wrongly, one of these malicious typos (i instead of 1), something that happens only in a seldom used if branch, something that looks perfectly ok when staring at it for hours but is still wrong. I know these kind of bugs, you can waste days finding them, and after finding them they suddendly become so obvious that you don't understand how you could overlook it.
I have been there, done that. A few times I have rewritten my code from scratch on the assumption that I won't make the same mistake twice. But of course that raises the real possiblity of making an entirely new error.
  • #9
  • Last Post: Nov 9, 2009 7:44am Nov 9, 2009 7:44am
Thank you for all your input. I have modified the program as 7 bit suggested, but (of course) everything is going as it is supposed to. I have gone over the code several times and all seems correct. So I don't get it...

I may end up rewriting it just to make sure its correct and there isn't something weird happening during the compiling process, although for the life of me I can't think of what it might be...I'll keep looking.

Thanks again
0 traders viewing now