Record the time when an indicator line crosses a threshold | Forex Factory

Options

Permalinks
Print Thread

Similar Threads

Draw line, count how many time price touches the line 4 replies

Need help- alert when trend line crosses price line 8 replies

Need simple indicator to record currency differentials real time in line 0 replies

Yesterday closing line and Today Opening Line Indicator 0 replies

Line crosses - Best Practices 1 reply

Record the time when an indicator line crosses a threshold

  • #1
  • First Post: Aug 4, 2010 12:24am Aug 4, 2010 12:24am
Anyone have an idea how to go about this? I'm trying to start the count from the crossing event.

I've looked at iTime and the closest I can think of is using a M1 array record and pulling the earliest time, but this seems a little overly complicated and the resolution is only to the closest minute.

Another idea I had was to create a far-off Pending Order when the event occurs and pull the time from the OrderOpenTime() but that's not so elegent.

Any coding gurus willing to shed some light?
  • #2
  • Aug 4, 2010 12:53am Aug 4, 2010 12:53am
Is there something wrong with storing the value in a static datetime variable?
That way you would have it accurate to the nearest second.

Here's some some code:

if (event) eventTime = TimeCurrent();

This assumes that only the most recent event time is of interest. To keep track of more than one would require an array.
  • #3
  • Edited 1:52am Aug 4, 2010 1:38am | Edited 1:52am
Thanks for the brainstorming, Codemeister.

How would you go about using TimeCurrent()? It returns the latest server time and not the instance of the time at the event.

I suppose it might be possible using Filewrite where the TimeCurrent() keeps getting recorded and the smallest figure is pulled. The whole set then purged when the condition becomes false. Might make a large csv though.

Anyone with a better workaround?
  • #4
  • Aug 4, 2010 1:52am Aug 4, 2010 1:52am
I thought the exact time didn't matter and than you were interested in the relative time. By that I mean if you know the time of the event you will alsway know how long ago it happened. If you had something else in mind, you need to explain it in detail.
  • #5
  • Aug 4, 2010 2:31am Aug 4, 2010 2:31am
Yes, I'm simply looking to record the time of the event.

Without the time of the event, I would not be able to subtract it from the TimeCurrent() to get the elapsed time.
  • #6
  • Aug 4, 2010 11:27am Aug 4, 2010 11:27am
I was tossing around the idea of using arguments with TimeCurrent() but no matter what, I will always get the running time and not the timestamp of the event.

Short of generating a far-off Pending Order and recording it's time, I'll can't think of any other method than to go the iTime route, which will only have the same resolution as the period of the indicators used.
  • #7
  • Aug 4, 2010 11:57am Aug 4, 2010 11:57am
Clearpaper,

Quote
I was tossing around the idea of using arguments with TimeCurrent() but no matter what, I will always get the running time and not the timestamp of the event.

I think you may have missed this part of CodeMeister's comment :-

Quote
storing the value in a static datetime variable?
  • #8
  • Aug 4, 2010 2:04pm Aug 4, 2010 2:04pm
Yes, so how would I go about storing the static datetime variable?

Say for example I want to record the time when a fast Moving Average crosses above a slow Moving Average.

double mafast_0=iMA(Symbol(),0,5,0,MODE_CLOSE,PRICE_MEDIAN,0);
double maslow_0=iMA(Symbol(),0,60,0,MODE_CLOSE,PRICE_MEDIAN,0);
if(mafast>maslow){
...then what?

Unlike an OrderSend event where we can just get the OrderOpenTime(), there's no timestamp.
  • #9
  • Aug 4, 2010 2:15pm Aug 4, 2010 2:15pm
Inserted Code
static int Time_Stamp;
double mafast_0=iMA(Symbol(),0,5,0,MODE_CL OSE,PRICE_MEDIAN,0);
double maslow_0=iMA(Symbol(),0,60,0,MODE_C LOSE,PRICE_MEDIAN,0);
double mafast_1=iMA(Symbol(),0,5,0,MODE_CL OSE,PRICE_MEDIAN,1);
double maslow_1=iMA(Symbol(),0,60,0,MODE_C LOSE,PRICE_MEDIAN,1);
if(mafast_0>maslow_0 && mafast_1<maslow_1)Time_Stamp = TimeCurrent();//timestamp at cross up
if(mafast_0<maslow_0 && mafast_1>maslow_1)Time_Stamp = TimeCurrent();//timestamp at cross down

I haven't tested it, but it should be something like this.
  • #10
  • Aug 4, 2010 2:21pm Aug 4, 2010 2:21pm
I'm not sure why capturing the time of the event is not considered a timestamp but those are only words.

The key to this is declaring a static datetime variable to store the most recent event time. Since the variable is static, it doesn't change and can be used to compute the elapsed time.

if (mafast>maslow) eventTime = TimeCurrent();
// you need to add code that ensures that this is only done once per event.

The elapsed time is easy to compute.
datetime elapsedTime = TimeCurrent() - eventTime;// in seconds
  • #11
  • Aug 4, 2010 2:56pm Aug 4, 2010 2:56pm
Actually, this would be better.........

Inserted Code
static int Time_Stamp;
static double mafast_1;
static double maslow_1;
double mafast_0=iMA(Symbol(),0,5,0,MODE_CL OSE,PRICE_MEDIAN,0);
double maslow_0=iMA(Symbol(),0,60,0,MODE_C LOSE,PRICE_MEDIAN,0);
if(mafast_1==0){
   mafast_1 = mafast_0;
   maslow_1 = maslow_0;
   return;
}
if(mafast_0>maslow_0 && mafast_1<maslow_1)Time_Stamp = TimeCurrent();//timestamp at cross up
if(mafast_0<maslow_0 && mafast_1>maslow_1)Time_Stamp = TimeCurrent();//timestamp at cross down
mafast_1 = mafast_0;
maslow_1 = maslow_0;
  • #12
  • Aug 5, 2010 12:47pm Aug 5, 2010 12:47pm
@CodeMeister: I think my misunderstanding stems from my noob ignorance of static variables. Laughable now.. Me:"Yes, I want to get a static snapshot." You:"Just use a static variable." Me:"???"

@kennyhubbard: Thanks for the code. I would have gone along the lines of your earlier example by defining mafast_1 and maslow_1 as the previous period, but your setting of them as the previous state in the last two lines of your second example is an elegant way. I would not have thought of that.

Is this part necessary though?
Inserted Code
if(mafast_1==0){  //
   mafast_1 = mafast_0;
   maslow_1 = maslow_0;
   return;
}
  • #13
  • Aug 5, 2010 12:56pm Aug 5, 2010 12:56pm
Hi Clearpaper,

Essentially what that bit of code does is prevent any errors on the first run of the EA(or producing of a false result).....now that I think of it, I cannot think of a good reason why the first method would not work, in which case you would not need that extra piece of code.
  • #14
  • Edited 2:31pm Aug 5, 2010 1:58pm | Edited 2:31pm
I see.. on starting the masfast_1 and maslow_1 would both be 0, and since both conditions require either mafast_1>maslow_1 or mafast_1<maslow_1, they would never fulfill.

But I suppose that's the beauty of the last two lines as they cover that initial state, right?

I used this and it worked
Inserted Code
   static int Time_Stamp;
   static double mafast_1;
   static double maslow_1;
   double mafast_0=iMA(Symbol(),0,5,0,MODE_SMA,PRICE_MEDIAN,0);
   double maslow_0=iMA(Symbol(),0,60,0,MODE_SMA,PRICE_MEDIAN,0);
   if(mafast_0>maslow_0 && mafast_1<maslow_1)Time_Stamp = TimeCurrent();//timestamp at cross up
   if(mafast_0<maslow_0 && mafast_1>maslow_1)Time_Stamp = TimeCurrent();//timestamp at cross down
   mafast_1 = mafast_0;
   maslow_1 = maslow_0;

And I think in the first example, the time recorded will be errorneous if the threshold is crossed more than once while within the same period. E.g.

Looking at the situation where the mafast_0 crosses above the maslow_0, therefore mafast_0>maslow_0.
The previous period's mafast_1<maslow_1, so therefore (mafast_0>maslow_0 && mafast_1<maslow_1)Time_Stamp = TimeCurrent()

If during the same current period, the mafast_0 crosses back below maslow_0, therefore mafast_0<maslow_0.
The previous period's ma's are still mafast_1<maslow_1, giving a situation of (mafast_0<maslow_0 && mafast_1<maslow_1) which condition is undefined.

The code does not capture the crossing down timestamp.

If the period is small, the effects would be negligible, but if the period is long, e.g. H4, the mafast_0 could cross over in the first hour, then cross back down in the last hour of the same period and the timestamp would be off by a difference of 2+ hours.

You see how sweet your second example is?
  • #15
  • Aug 5, 2010 2:11pm Aug 5, 2010 2:11pm
Exactly :-)

In hindsight, the extra code for the first run was never required since 0<0 will always return false so it won't take a timestamp. It does serve as a good lesson though for the future, in that you always need to keep an eye on the first run of a static variable since there may be cases where it will produce true (eg Ask > maslow_1). This would produce a false positive on the first run.
  • #16
  • Aug 5, 2010 2:45pm Aug 5, 2010 2:45pm
Yes, it most certainly is a good and interesting lesson. Thank you!

I was editing my previous post while your reply came in, I made some observations of your 1st example and how I think the second one solves a big problem in the 1st. Might be wrong, your thoughts would be appreciated.
  • #17
  • Last Post: Aug 5, 2010 2:56pm Aug 5, 2010 2:56pm
Yup, you are correct in your observations.......

I thought there may be a problem taking the reference for the crossing off the previous bar, rather than off the previous tick.
0 traders viewing now