Defining Trend | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Defining market conditions 5 replies

Defining who is a better trader... 47 replies

Defining a trend?! 2 replies

Defining shade 1 reply

Attachments: Defining Trend
Exit Attachments

Defining Trend

  • #1
  • First Post: Feb 26, 2010 10:02pm Feb 26, 2010 10:02pm
Hi All,

I have been looking for a good and simple trend code for entry that would see trend just as I would see it if I was looking at the chart.

I have put my idea that is coded below and when I manually sit and try it on the charts it seems excellent and the trend updates continually unless there was a fractal in the last 30 minutes that it uses instead.

If someone would put this in an indicator to draw lines to check it, great. Else it might be useful for someone as is, in an EA


This is how it works.. take last two fractals that are both above or below the bollenger band and find the delta. Take that point and draw a line to the bollenger band upper or lower (must be on the same side as the two fractals) and that is the trend.

Inserted Code
 
      double BuyLine = iBands(NULL,15,20,2,0,PRICE_LOW,MODE_LOWER,0);
      double SellLine = iBands(NULL,15,20,2,0,PRICE_LOW,MODE_UPPER,0);
      

//Define Trend

      int Trend;
      static int HighFractalTime, LowFractalTime; 
      static double FirstHighFractal, LastHighFractal, HighFractal,FirstLowFractal, LowFractal, LastLowFractal;
     
      if (iFractals (NULL,15, MODE_UPPER, 0) >= SellLine) // Only count fractal greater than upper bollenger band 
      {FirstHighFractal = LastHighFractal; LastHighFractal=HighFractal;  HighFractal= iFractals (NULL,15, MODE_UPPER, 0); HighFractalTime=Time[0];}
      if (iFractals (NULL,15, MODE_LOWER, 0) <= BuyLine) // Only count fractal lower than lower bollenger band 
      {FirstLowFractal= LastLowFractal; LastLowFractal=LowFractal; LowFractal= iFractals (NULL,15, MODE_LOWER, 0); LowFractalTime=Time[0];}

      if (HighFractalTime < LowFractalTime) // Last fractal was Low
      {      
      if (( TimeCurrent() - LowFractalTime )/60 > 30) { 
         if((LastLowFractal+LowFractal)/2 > BuyLine + 200*Point) Trend=1; //Trend is Down
         if((LastLowFractal+LowFractal)/2 < BuyLine - 200* Point) Trend=2; //Trend is Up
         if (Trend != (1||2))Trend=3; //Trend is Ranging
         }
      if (( TimeCurrent() - HighFractalTime )/60 <= 30) {
         if((FirstLowFractal+LastLowFractal)/2 > LowFractal + 200 * Point) Trend=1;  //Trend is Down
         if((FirstLowFractal+LastLowFractal)/2 < LowFractal - 200 * Point) Trend=2;    //Trend is Up
         if (Trend != (1||2))Trend=3;//Trend is Ranging
         }}
      
      if (HighFractalTime > LowFractalTime)// Last fractal was High
      { 
      if (( TimeCurrent() - HighFractalTime )/60 > 30) { 
         if((LastHighFractal+HighFractal)/2 > SellLine + 200*Point) Trend=1; //Trend is Down
         if((LastHighFractal+HighFractal)/2 < SellLine - 200* Point) Trend=2; //Trend is Up
         if (Trend != (1||2))Trend=3; //Trend is Ranging
         }
      if (( TimeCurrent() - HighFractalTime )/60 <= 30) {
         if((FirstHighFractal+LastHighFractal)/2 > HighFractal + 200 * Point) Trend=1;  //Trend is Down
         if((FirstHighFractal+LastHighFractal)/2 < HighFractal- 200 * Point) Trend=2;    //Trend is Up
         if (Trend != (1||2))Trend=3;//Trend is Ranging
         }}
Any suggestions on code improvement would be very welcome. I am no great coder.. so there may be a lot better way to do it.

cheers
A
  • #2
  • Feb 26, 2010 10:33pm Feb 26, 2010 10:33pm
hey asgard..... if you have the time, had a question about a phrase in your code......

Inserted Code
 
if(Trend != (1||2))

does that really work..... != (1||2) ...... or will it always be false......h
to trade and code, keep both simple... no call to impress....h
  • #3
  • Feb 26, 2010 10:44pm Feb 26, 2010 10:44pm
Quoting hayseed
hey asgard..... if you have the time, had a question about a phrase in your code......

Inserted Code
 
if(Trend != (1||2))
does that really work..... != (1||2) ...... or will it always be false......h
Actually, I didn't use it in the EA yet, because I am still writing it.

I thought it should work but I might be wrong.

I do have a mistake in the code though.. the last fractal to use as an end point does not need to be above the bollenger band.

I will fix it and see if that code works that you mention
  • #4
  • Feb 26, 2010 10:49pm Feb 26, 2010 10:49pm
hey asgard..... i believe you'll have to break it into 2 parts..... or at least that has been my experience......

the best method for trend determination has been a long quest of mine also......h
to trade and code, keep both simple... no call to impress....h
  • #5
  • Feb 27, 2010 2:52am Feb 27, 2010 2:52am
Yes Hayseed, I think you are right about that code.. needs to be separate statements.

I tested this about this time last night but I had not written the code at the time. I had someway of filtering the bollenger band delta on a large fall or rise.

I am missing something so I need to go through my test notes and update the code in the morning.

cheers A
  • #6
  • Feb 27, 2010 4:03am Feb 27, 2010 4:03am
hi asgard,

first of all, thanks for sharing your idea.
to define the trend in an ea was always a hard part of the code for me, but this one seems to be a very good one

could you please attach a picture for better understanding

thanks,
-ba
  • #7
  • Feb 28, 2010 5:07am Feb 28, 2010 5:07am
I too would like to see the idea on a chart so I can better understand it.

I'm not sure why you limited the code to 15M TF. It seems it would work on the higher TFs as well and it might even be possible to use it in a Multiple Time Frame strategy.
  • #8
  • Edited 11:46am Feb 28, 2010 10:23am | Edited 11:46am
Ok, this took me a little longer than I thought and I still haven't done the code

I changed the endpoint price to the bottom envelope instead of the bollenger band.

The blue dots are the anchor points created by the fractal division.

The arrow is the angle it sees at the arrowhead time. The arrow head will follow the envelop line continually while the anchor point will stay the same until the next fractal.

I picked a volatile time to see what the result would be and it seems like it would be ok to enter trades based on most of it but it has missed the ranging market because of the previous volatility. It would not do that if I had changed to the top fractals after the sudden rise.


Needs to be a rule for the fractals if there is not one in say 12 hours or when you get two within the space of a few bars.

I will code this now.. post it here soon.

cheers A
Attached Image (click to enlarge)
Click to Enlarge Name: chart3.jpg Size: 127 KB
  • #9
  • Feb 28, 2010 11:22am Feb 28, 2010 11:22am
Quoting asgard2
Hi All,

I have been looking for a good and simple trend code for entry that would see trend just as I would see it if I was looking at the chart....
This function will give you the trend of a currency pair. Just call it with the pair and timeframe.

Scoobs.

//+------------------------------------------------------------------+
//| expert Trend_Strength function |
//+------------------------------------------------------------------+

string Trend_Strength(string symbol, int tf)
{

string Trend = "F";

double ADX01 = iADX(symbol,tf,10,PRICE_CLOSE,MODE_MAIN,0);

double EMA01 = iMA(symbol,tf,025,0,MODE_EMA,PRICE_CLOSE,0);
double EMA02 = iMA(symbol,tf,050,0,MODE_EMA,PRICE_CLOSE,0);
double EMA03 = iMA(symbol,tf,100,0,MODE_EMA,PRICE_CLOSE,0);

if(ADX01 > 20)
{
if(EMA01 > EMA02 && EMA02 > EMA03)
{
Trend = "U";
}
if(EMA01 < EMA02 && EMA02 < EMA03)
{
Trend = "D";
}
}

return(Trend);

}
  • #10
  • Feb 28, 2010 11:49am Feb 28, 2010 11:49am
Quoting scooby-doo
This function will give you the trend of a currency pair. Just call it with the pair and timeframe.
Thanks Scoob

The trouble with moving averages is they don't filter noise unless you make them a long period. Then they take too long to respond to reversals.

For me, I wanted something different.
  • #11
  • Feb 28, 2010 12:00pm Feb 28, 2010 12:00pm
Quoting CodeMeister
I too would like to see the idea on a chart so I can better understand it.

I'm not sure why you limited the code to 15M TF. It seems it would work on the higher TFs as well and it might even be possible to use it in a Multiple Time Frame strategy.
You're right CodeMeister, it need not be limited to 15 minutes and could be used instead of or in addition to another time frame.
  • #12
  • Feb 28, 2010 7:01pm Feb 28, 2010 7:01pm
The chart makes it clear how the thing works. It seems from your comments, this is still a work in progress. The MA approach is what I have been using and I have been looking for a better one. I'm not convinced yet that this is the answer but I will follow the thread.

I would like to see someone with a strong mathmetical background make a comment about this.
  • #13
  • Edited Mar 1, 2010 12:22am Feb 28, 2010 11:40pm | Edited Mar 1, 2010 12:22am
Quoting CodeMeister
The chart makes it clear how the thing works. It seems from your comments, this is still a work in progress. The MA approach is what I have been using and I have been looking for a better one. I'm not convinced yet that this is the answer but I will follow the thread.

I would like to see someone with a strong mathmetical background make a comment about this.
Great thoughts.. I am testing it live manually and so far really good (better than that by results). My problem is coding it to enable testing over a long term. I am still trying to get my head around the detail for coding it.

Updated Pic because I missed a fractal point below the line.

The other problem is that the trend cant be seen in an EA or I don't know how to make that happen

Example of the problems.. say you have a significant fractal but it misses the line by Just one point. Manually you would take it as a hit but the computer will not see it that way.

Or say you see a spike of 40 points above the line, manually you would account for that and include the point but at a lower value.

The theory is simple enough, I dont know for sure, that it always works. It seems to and it's as close as I have managed to get so far and it is definitely more accurate and responsive than an MA delta or cross or both.

In my opinion, there is no way to draw a trend without an anchor point, if the market changes the anchor point should stay as a reference.
Attached Image (click to enlarge)
Click to Enlarge Name: chart4.jpg Size: 142 KB
  • #14
  • Mar 24, 2010 6:53pm Mar 24, 2010 6:53pm
This thread is what I meant by my comment about having a solid mathmetical basis for the indicator. Asgard2, have you abandoned this idea or still working on it?

http://www.forexfactory.com/showthread.php?t=225112
  • #15
  • Mar 26, 2010 7:39pm Mar 26, 2010 7:39pm
Hi CodeMeister,

I did not abandon the idea but it became clear really quickly, that no one has the same idea of what a trend is.

Actually on reflection, I don't think by my definition, there is a single trend, I think there are at least three. And I think occasionally they can be time based as easily as they can be price based. The longer duration of the trend has a probability increasing to the upside as the trend continues.

I came across the age old problem of what is too much detail and when is a calculation nothing more than a hypothesis that has little bearing in the real world.

So out of the three trends (long term,medium,short), I decided to put a value on each depending on the trade type and run a probability calculation for the input of the signal based on candle patterns and where the price is based on fibo line and where the price has come from.

That gives me a ballpark entry price. Then I look in more detail at what the last few bars have done and where they are and which market is open.
It is taking a while to do (7 Months now) and when I get it done, I will post it so people can try it.
  • #16
  • Edited 7:56pm Mar 26, 2010 7:55pm | Edited 7:56pm
OK, in the meantime here is another thread devoted to this subject. Its encouraging to see that so many people are interested in finding a solution to this seemingly simple problem.

http://www.forexfactory.com/showthread.php?t=227948

I have been trying for 3 or 4 months myself and have made some progress, but I am still not satisfied that it worth posting. In my opinion its better than what has been posted, but the mathmetical basis is not solid and I am sure that with some more inspiration, I can come with something better. If you find something better than mine, I would definitely be interested. In the meanwhile, I am going to try the other threads' ideas.
  • #17
  • Mar 26, 2010 8:03pm Mar 26, 2010 8:03pm
Codemiester,

for some reason I cannot post in that thread that you linked to but this is what I think. http://www.forexfactory.com/showthread.php?t=225112

As far as price agreements go that he mentions.. they are not trend changes except at tick level an cannot be measured to use a future reference.

if a trend does not have an anchor point relative to price.. what is a spike and is that a new price level agreement?

Not trying to be rude but it looks to me like some pretty nice spin on a flawed idea.

Quote
If I had to give a dictionary definition of trend it would go something like this.

A trend is, steady market momentum in the one direction that is determined only by past and future events. Each time the momentum ceases to maintain a consistent vector the trend can be considered to be altered.
  • #18
  • Mar 26, 2010 8:26pm Mar 26, 2010 8:26pm
I think the thread originator put restrictions on his thread. I can't post either. The restrictions of little good - some of the usual suspects have shown up already.

I share your opinion (I think). He strikes me as one of those academics who will defend his ideas to the last breath even if it is useless. I'm not interested in what works in the classroom, I want something that works in the real world, but has some mathmetical backing so I don't have to test it exhaustively and then defend myself from the naysayers.

My latest idea is not one calculation or formula, but to combine several and weight them to come up with the answer. So any new ideas that look worthwhile might fit well into my scheme and that's why I am interested.
  • #19
  • Mar 26, 2010 9:07pm Mar 26, 2010 9:07pm
Quoting CodeMeister

My latest idea is not one calculation or formula, but to combine several and weight them to come up with the answer. So any new ideas that look worthwhile might fit well into my scheme and that's why I am interested.
I agree, the way to get the most consistent results is to determine what leads to the mathematical probability of a trend change. Leads to the next question of which trend.

On a simpler note

step 1 is determine the past trend/s along with the momentum/time.

step 2 evaluate what other traders see or will do. (fibo, 50% rule and leg rules) What do they see as a change in trend rather than a retracement area?

step 3 detect the shift in momentum and give it a value. (Micro bars and tick power) when professional money is in the market you can see it in the tick charts... market will go one way slowly and the last 5 slow ticks in one direction will be taken out by sudden jumps in the other, followed by more slow ticks.

If it is getting to session close time for say London.. if they haven't managed to change all the money they need to, sometimes they will just dump the rest in a counter trend that might last just a couple of bars before a breakout in the other direction.

All of these steps need to be broken down and given a value of importance. When a value is right, in theory the chance of the trade working out will have far greater odds.

are you an aussie?
  • #20
  • Mar 26, 2010 10:45pm Mar 26, 2010 10:45pm
Quoting asgard2
are you an aussie?
Canuck.

I guess the natural progression after determining the trend is some measurement of momentum, but you have introduced some considerations that go further than I have seen or thought about. What you are proposing is more of a system than a simple definition of trend. I was thinking more of something simple I could add to my existing systems to make them better.

You have definitely given me something to think about.
0 traders viewing now