Visual Interpretation to Mathematical Values Problem | Forex Factory

Options

Permalinks
Print Thread

Similar Threads

Need Help And Explanation On Interpretation Of News Report 9 replies

COT Interpretation 3 replies

Visual backtest result different from non-visual backtest 0 replies

Trailing Stop Expectancy - A Mathematical Problem 41 replies

Visual Interpretation to Mathematical Values Problem

  • #1
  • First Post: Jul 13, 2010 3:47am Jul 13, 2010 3:47am
Hello,

I really need some advice regarding a programming problem I'm experiencing. I cannot really advance any of my EA's without solving this problem.

The problem lies in converting any visual analysis into code. For instance, look at the picture below, a simple 34EMA line. The eye can see when the line flats out to enter a ranging period, and the eye can easily see if the line trends upwards for an upwards trend. But how do you program code to do the same?

The problems that I have stumbled upon:

 

  1. The trend line, as the human eye sees it, is based on the resolution of the screen.
  2. You can calculate static empirical values (number ranges where you attach a specific market trend to it), but there is no guarantee that they will work in the future because those numbers are calculated based on the now and past market data.
  3. Each timeframe has different values. You cannot do an empirical test, and use those static values for all timeframes.
  4. There isn't really a correlation between the timeframes, if there is one it is relative to the market condition.


Those are the main problems that I have stumbled upon.

If anyone could help me regarding this issue I would really appreciate it.

Regards

http://img375.imageshack.us/img375/745/88323471.gif

  • #2
  • Jul 13, 2010 3:52am Jul 13, 2010 3:52am
I should also add that the closest I came to making this work is to do the following:

Calculate the values for the past 2 years or any long period of time.

Divide that values by the value of a point, or the Point function(so that the same formula and parameters can be used on all currency pairs).

Multiply that average by certain percentages to construct ranges for certain market periods.

If anyone can improve on this logic please contribute.

Regards
  • #3
  • Jul 13, 2010 7:44pm Jul 13, 2010 7:44pm
I spent considerable time working on much the same problem and didn't solve it. I tried others code such as Heiken Ashi and HMA but never really came up with something that didn' whipsaw during ranging conditions.

I think the ultimate solution will consist of something that combines more than one indicator and each indicator would be set for multiple periods so that you could measure the short, medium and long term trend.

But as many comments on this forum have pointed out - there is nothing that will beat an experienced pair of eyes. If you come up with something even close, you will probably have found the Holy Grail.
  • #4
  • Jul 14, 2010 4:57am Jul 14, 2010 4:57am
So far this is the best that I could do for a moving average:

Inserted Code
double MA_Avg_Trend(int tf,int period,int bars)
{
double MA[];
double sum = 0;
int count;

ArrayResize(MA,bars+1);


for (int i =0;i<=bars;i++) { MA[i] = iMA(Symbol(), tf, period, 0, MODE_EMA, PRICE_MEDIAN, i)/Point; }

for (i =0;i<bars;i++) 
{
if ( MA[i] != 0 && MA[i+1] != 0 )
{
sum = sum + MathAbs(MA[i] - MA[i+1]);
count++;
}
}

return(sum/count);

}


Now that you have the average trend value for the past x bars, you can compare the current trend value with the average historic trend values, and thus make a decision if the trend is indeed strong or not according to the moving average.

If anyone can improve on this method please do so!
  • #5
  • Last Post: Jul 14, 2010 11:29am Jul 14, 2010 11:29am
Quoting Saidar
Hello,

I really need some advice regarding a programming problem I'm experiencing....

You're not making yourself very clear here but for the record

 

  1. There isn't really a correlation between the timeframes, if there is one it is relative to the market condition.

There are two constants over all timeframes : Price and Time
MAs on one frame may have a different values on another but if you take the close price of two hourly bars say at 1am and 9am you have 8 hours time elapsed.
The same difference in price and time elapsed would be true on a 1 minute chart.


With this you can calculate a similar angle of attack which remains the same on any chart

pipdifference = (priceend - pricestart)

if pipdifference <0 then down trend
angle = -90 + (1/TAN(Abs(pipdifference)/timeinmins))


if pipdifference > 0 then uptrend
angle = 90 - (1/TAN(Abs(pipdifference)/timeinmins))


where 90 and -90 are vertical up and down trends and approach 0 / -0 for ranges.

The fact that the same value MA on different timeframes will give different values is what you want isnt it ? After all you can have a strong 1 minute uptrend whilst being in a strong hourly down trend.

You then only need to work out (or best guess) what is the best period (in minutes) for each timeframe to calculate the angle over.

EG hourly = 8 bars = 480 minutes
15 min = 14 bars = 210mins

etc.

Now its another thing entirely how you gauge which timeframe to weight over another

EG Hourly = -45degrees 5 min = 60degrees

Is the hourly trend still in effect or is it ending ??

Anyway at least with a solid method of calculating an angle of attack you may be able to make some progress

HTH

0 traders viewing now