ATR = Exponential Average of the True Range ?

[aaa 10]

Hello

 

I looked inside the function AvgTrueRange

 

AvgTrueRange = Average( TrueRange, Len ) ;

 

I thought that The AvgTrueRange was the Exponential Average of the True Range

 

So it should be

 

AvgTrueRange = XAverage( TrueRange, Len ) ;

 

Anyone can tell which average to use ?

[zdo 20]

Your choice...depends on your needs. I doubt if ATR application originators (LeBeau, etc) have made one way 'official'

 

Average will give you the central tendency for all the ranges in the sample (length)

xAverage will give more weight to more recent ranges in the sample (length)

 

An exceptionally large recent range will push ATR number higher with xAvg than with Avg...

A sample with an exceptionally large range in it will stay up longer with Avg than with xAvg as the large range fades to the bottom of the sample (length)

 

Long length with Avg works pretty well if you're doing any std dev work

Shorter length with xAvg works pretty well if you need higher sensitivity to recent activity...etc

 

Which one suits your needs better?

[aaa 10]

ThanX zdo

 

They is an other way to compute the ATR Grrrrrrr...

 

ATR.TrueRange = ( 9 * ATR.TrueRange-1 + TrueRange ) / 10 ;

 

if we look at the pix below the 3 are very differents !!

 

 

I'm suprised that there is no "official" ATR

 

I don't look to tweak it only to know the most popular one

 

I try to stay basic

 

If TS or MC (I've checked 2 others soft) uses Average they will be a lot of traders in this way !?

 

Is anyone uses xAverage for ATR ?

 

[Tams 20]

you can plot all three lines on the chart and see which makes more sense.

 

 

 

 

p.s. nice print out !

Edited October 10, 2009 by Tams

[aaa 10]

p.s. nice print out !

 

Coz I've started with a nice teacher...

 

When it's bad Tams says it...

 

And when it's good ,Tams also says it.

 

But I'm always happy with both when it comes from my TamTams

 

Have a nice WE

 

PS I've an other idea...

[aaa 10]

Even tho I don't believe in backtesting I post an ATR strategy with the 3 types of ATR

 

As we see arithmetic looks to be the best choice

 

[Tams 20]

I like that nA, nX and nTR.

That's clever !

 

;-)

[NetTecture 10]

With all respect...

 

an exponential moving average would NOT result in an ATR. Period. The "Average" is mathematically well defined. Same with a normal moving average - which is why the exponential moving average has a different name.

 

Assuming the ATR is not a normal average is a little - sorry - dump. It says clearly it is the AVERAGE of the true range, which definitely IS officially defined.

[sevensa 10]

With all respect...

 

an exponential moving average would NOT result in an ATR. Period. The "Average" is mathematically well defined. Same with a normal moving average - which is why the exponential moving average has a different name.

 

Assuming the ATR is not a normal average is a little - sorry - dump. It says clearly it is the AVERAGE of the true range, which definitely IS officially defined.

 

Where do you get your official definition of Average from that Average means Normal Average? According to Wikipedia, Average can mean different things and not Normal Average when someone say Average. What source did you use?

 

Btw, what does "a little dump mean"? Is that when someone think dump is the same as dumb or that their assumptions make something official?

[NetTecture 10]

It is called brain usage. Any non-normal average says so. THis is why the "moving average" is the normal one, and guess what - the "exponential moving average" is - the exponential one.

 

Normal means normal. Anything else is not normal.

 

Some people think too complex and fail to see the obvious, and obviously the OP belongs into that category.

[sevensa 10]

It is called brain usage. Any non-normal average says so. THis is why the "moving average" is the normal one, and guess what - the "exponential moving average" is - the exponential one.

 

Normal means normal. Anything else is not normal.

 

Some people think too complex and fail to see the obvious, and obviously the OP belongs into that category.

 

Please point out a source saying that Moving Average means Normal Moving average as in your statement I have marked in bold above. The definition says Average, not Normal Average. I have referred you to a source showing that Average can mean different things. All you have is your assumption that Average means Normal average and call anyone who disagree with your assumption dumb. (errr.. dump). "Brain usage" is not a valid reason to prove something.

[aaa 10]

Hello

 

I've changed Data ranged from 2500 bars to 2550 bars

 

Surprise !

 

The result is different

 

Then from 2500 to 2450

 

The same surprise

 

-my strategiet is not robust ?

 

-Backtesting is not the way to this question ?

 

-Can we trust Backtesting ?

[Tams 20]

....-Can we trust Backtesting ?

 

 

well, have you heard of the term GIGO ?

 

Garbage In, Garbage Out...

 

 

for useable backtesting results... you must start with good data

 

good data means more than data integrity.

 

of course you have to start with data that are clean -- i.e. with no holes or bad ticks in them.

 

good data also means that the sample is a representation of the population.

 

e.g. if your data range is between Sept 08 to Nov 08, you will see a very biased downtrend in the market,

any trend following system, especially down biased system, will do well.

But the same system might fall flat on the face when the sample data range is changed.

 

 

 

these are just some thoughts to get the thinking going...

[aaa 10]

...

 

these are just some thoughts to get the thinking going...

 

Dear Tams

 

Everybody would be happy to share thoughts with someone

 

So I am !

 

If the question is concerning enough peole, there is always here someone to share thoughts with

 

This is why I like this site

 

It's like a big brain

 

-----------------------------------------------------------------------------

 

well, have you heard of the term GIGO ?

 

Yes sir !

 

In French with a T at the end, It is a very good part of the lamb to eat ;=)

 

Seriously, My Data source are the 30 stocks of the DJ offLine

 

Is it A good representation of the population.?

 

I've checked that they are no holes in

 

-----------------------------------------------------------------------------------------------

 

e.g. if your data range is between Sept 08 to Nov 08, you will see a very biased downtrend in the market,

any trend following system, especially down biased system, will do well.

But the same system might fall flat on the face when the sample data range is changed.

 

I've used 2500 bars (10 years of test )

 

How Changing only +- 50 bars/days days could change so much the result ?

 

-------------------------------------------------------------------

 

Some Other thoughts :

 

Hazard effects

 

Optimization paradoxe

 

Overfitting

 

 

Kind Regards

 

aaa

Edited October 27, 2009 by aaa

[Tams 20]

...

e.g. if your data range is between Sept 08 to Nov 08, you will see a very biased downtrend in the market,

any trend following system, especially down biased system, will do well.

But the same system might fall flat on the face when the sample data range is changed.

 

I've used 2500 bars (10 years of test )

 

How Changing only +- 50 bars/days days could change so much the result ?

...

 

 

if you can pin point the 50 bars that made the difference...

all you need is to hand calculate the variables to find out the pivotal factor.

 

 

 

Seriously, My Data source are the 30 stocks of the DJ offLine

 

Is it A good representation of the population.?

 

I've checked that they are no holes in

 

 

are you talking about a portfolio back testing of 30 stocks?

the permutation can be exponential.

[aaa 10]

all you need is to hand calculate the variables to find out the pivotal factor.

 

very interesting

 

are you talking about a portfolio back testing of 30 stocks?

the permutation can be exponential.

 

No I 've tested difeerent stocks from the DJ with different period

 

I don't feel confortable with strategie = I'm not sure if it is realistic

 

This is maybe the reason why it influenced with bad vibrations my test !!!

 

Because I'm curious I will certainly ask more question in an other topic about strategie

 

Anyway I choose the arythmetic average for my projects

 

MANY THANKS TO ALL

[Tams 20]

I would use the PRINT statement to extract the variables during the backtest...

to see which variable that is causing the unfavored impact.

 

 

this indicator might help,

I have not tried it, but it looks useful.

 

http://www.traderslaboratory.com/forums/f46/open-equity-curve-indicator-6943.html#post77607

 

[aaa 10]

I would use the PRINT statement to extract the variables

 

I've done it

 

With your precious indications from your excellent thread about PRINT

 

I always use sophisticated Print for debug + To file

 

It's now a part of all my codes

 

----------------------------------------------------------------------------------------------

 

I know this indicator Tams

 

I'm the one whom has posted it

 

But I didn't arrived to make it working with MC

 

Tooooo bad it looks interesting

 

 

{== START OF HEADER ==========================================================================

Program:  Open Equity Curve Indicator
Date:     July 2009
Platform: TradeStation v8.5

DESCRIPTION: Display a strategy's open equity curve.

DEVELOPMENT NOTES:

08/02/09 Initial version.

TO-DO LIST:

== END OF HEADER =============================================================================

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Copyright ? 2009. Capital Evolution, LLC. All Rights Reserved.
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

== DEFINE ALL INPUTS AND VARIABLES ==========================================================}

Input: PositiveColor( Green ),
   NegitiveColor( Red ),
   ZeroLineColor( White );

Variables: OpenEq(0);

OpenEq = I_OpenEquity - I_ClosedEquity;
If ( OpenEq > 0 ) Then SetPlotColor( 1, PositiveColor )
Else SetPlotColor( 1, NegitiveColor );
Plot1( OpenEq , "Open P&L" ) ;
Plot2( 0, "Zero Line", ZeroLineColor );

{== END OF MAIN PROGRAM =====================================================================
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Copyright ? 2009. Capital Evolution, LLC. All Rights Reserved.
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++}

Edited October 28, 2009 by aaa