Everything about RSI - page 54

 

Thanks A lot!!

 

Looks great! but it is possible to add the histogram??

Can't thank you enough for your help.

Files:
gbpusddaily.png  33 kb
 
gorangel:
Looks great! but it is possible to add the histogram?? Can't thank you enough for your help.

gorangel

Here you go

 

Perfect!! thanks a lot!!

Files:
 

Hi, Can you do it with alarm,please!

4 time frame RSX.mq4

 
mladen:
And this one is showing what can be done when the calculation method of rsi from above is used and when averages are replaced with smoother. It gives results very close to rsx (as it is obvious) but it is still not exactly the same as rsx (due to a fact that rsx uses one specific way of smoothing). Difference in results is very small so we could even call this one a kind of an rsx, but for the sake of less confusion, I named it a rsi of a smoother 2.

So, all in all, knowing how something is calculated, gives us opportunity to actually make a completely new indicator, and that is what Mark Jurik exactly did : he took an rsi and replaced SMMA with a custom smoothing and named that RSX. We can make that way as much rsi types as we have smoothing or averaging ways, the list is endless.

So this one concludes the "how" can some new indicator be made and how can a long existing indicators be improved. This version is much smoother than the original rsi (less false signals) without any lag added to it which automatically makes it a better candidate for using in trading then the original RSI

Hi Mladen,

I ususally catch in your logical coding structure this phrase :

if (A!=0) then B = (.../A)

else then B = ....

I wonder why you don't replace that by adding a universal epsilon error to A (right in part of parameter setting or indicator's initialize) to avoid the case of zero denomiator. By this tiny addition, your indicator and coding structure would be smoother (in looking ), and focus more on its useful functions.

Yours respectfully,

fareastol

 
fareastol:
Hi Mladen,

I ususally catch in your logical coding structure this phrase :

if (A!=0) then B = (.../A)

else then B = ....

I wonder why you don't replace that by adding a universal epsilon error to A (right in part of parameter setting or indicator's initialize) to avoid the case of zero denomiator. By this tiny addition, your indicator and coding structure would be smoother (in looking ), and focus more on its useful functions.

Yours respectfully,

fareastol

fareastol

If I understand you correctly you mean why don't I use some tiny value instead of that control.

Two reasons : because a result of some operation can be 0 (and then initializing the variable that will receive a result of that operation to some non zero value would be of no use) and because it would create a bias (either + or - values would be "preferred" and, in some cases, when the values used for calculation are small, it could create a huge deviation from a correct value). And the third (untold reason) : mathematically division by 0 is undefined value not some small or large value. I simply like to control the case when it is undefined

 
mladen:
fareastol

If I understand you correctly you mean why don't I use some tiny value instead of that control.

Two reasons : because a result of some operation can be 0 (and then initializing the variable that will receive a result of that operation to some non zero value would be of no use) and because it would create a bias (either + or - values would be "preferred" and, in some cases, when the values used for calculation are small, it could create a huge deviation from a correct value). And the third (untold reason) : mathematically division by 0 is undefined value not some small or large value. I simply like to control the case when it is undefined

You're definitely right, Mladen ! Thanks to your explanation, i recognize that idea of epsilon may cause big problems in calculating and precise results then.

Happy weekend to you.

 

Mladen, look, please, indicator Unsmoothed RSI, it sometimes starts to show incorrectly and it is necessary to switch ТF that readings of the indicator were corrected

 
Riksha:
Mladen, look, please, indicator Unsmoothed RSI, it sometimes starts to show incorrectly and it is necessary to switch ТF that readings of the indicator were corrected

Try this one out. As far as I see this version should not have any problems

Files: