Registered User
Joined: 3/21/2006
Posts: 19
|
TC2000 and Stockcharts cakculates the above indicator differently. I recently read that TC2000 is calculating this indicaator incorrectly. Stkchart calculates the ATR for DJ-30 as of 12/2 to be 221.76 and TC200 shows 206.98-big difference. See page 23 of
Wilder's book "New Concepts in Technical Trading Sstems" to get the correct formula.
Just passing on to you what I read-might want to check this out.
Jimmy H Joyner
|
Registered User
Joined: 7/30/2007
Posts: 1,072
|
I'm not currently using ATR, but I saw that same write-up on StockCharts' blog. I've got Wilder's book, but I haven't taken the time yet to calculate the ATR myself in a spreadhsheet so I can compare the results (the Bears kick off in less than an hour).
Here's an excerpt of the blog entry. Hopefully Bruce can shed some light on the different calculations (I'm not saying one is right and the other wrong - I'm just curious about the different methods to arrive at an ATR) ...
AVERAGE TRUE RANGE COMPARISON
The other day, someone wrote in and asked why our ChartSchool formula for Average True Range (ATR) was different from what they saw on other websites. Curious, we looked into the issue and were very surprised to discover that several other websites were describing ATR incorrectly.
ATR was created by J. Welles Wilder - the creator of the RSI indicator, the ADX indicator, the Parabolic SAR indicator, and several others. He first described it in his amazing book "New Concepts in Technical Trading Systems."
The first part of the equation involved computing something called the "True Range." Everyone gets that part correct. (You can see the details for yourself here.) It's the second part - the "Average" part - that many other websites are getting wrong.
On page 23 of Wilder's book it says:
"The equation for the AVERAGE TRUE RANGE is as follows:
ATR(latest) = ( 6 x ATR(previous) + Today's True Range ) / 7
To get the ATR initially, add the true range, as defined, for the past seven days and divide by seven. The answer from this will be used as the ATR(p) in the equation for the next day."
On page 26, Wilder reiterates the steps:
"2. ATR -- AVERAGE TRUE RANGE
A. Initially obtained by adding the true ranges for seven days and dividing by seven.
B. The latest ATR is obtained by multiplying the previous ATR by six, adding today's true range and then dividing the total by seven."
Those discussions are in the context of creating a 7-day ATR. To create the more typical 14-day ATR, you would use 13 instead of 6 and 14 instead of 7 in the above formula. That's the same as the information we present in our ChartSchool article and it's what we use in our programs.
Unfortunately, other website just take the 14-period Simple Moving Average of the True Range values and call that average the ATR. That is a huge mistake on their part caused - we guess - by laziness.
To see why, lets look at the current ATR value of the Dow Jones Industrial Average using our website and a popular competitor's site:
So we show that the Dow's ATR(14) should be 221.762 whereas our competition shows it at 206.98. That's a HUGE difference! Why is that?
The answer is obvious if you look at the following Excel spreadsheet that calculates ATR using both Welles Wilder's way of averaging things and the simple moving average method. Here's a screenshot of the bottom of that speadsheet:
And there you go - If you use the Wilder averaging method you get 221.76 which is what our chart shows. If you don't do your homework and assume that "Average" means "Simple Moving Average", you get 206.98 just like our competitor does.
(BTW, our chart shows the 200-day EMA at 11695.15 whereas the competition has it at 11703.14. Here a link to an article explaining why our number is correct in that case as well.)
The bottom line here is that StockCharts.com is 100% dedicated to accurate calculations and we take the time to go to the authoritative source for our formulas. Not everyone does that and, if they can't get ATR correct, what else do they have wrong?
|
Registered User
Joined: 1/28/2005
Posts: 6,049
|
As I understand it, True Range (the individual calculation from each bar) is always the same.
When you put together an average over your time period (ATR) you need to select a smoothing method:
simple,
exponential,
Wilders (the smoothest)
are most popular.
The smoothing selected will cause the differences.
(much like using a simple, exponential, or a weighted moving average)
There’s really no “wrong” calculation its just a matter of what’s used.
Thanks
diceman
|
Registered User
Joined: 7/30/2007
Posts: 1,072
|
Thanks Diceman! .... makes sense.
Like I said, I don't currently use ATR, but since Wilder originally developed it, and since your post indicates that it's the "smoothest smoothing method" (try saying that several times really fast) it's curious to me that that is not the smoothing method used here.
At the very least, since it is a pretty popular indicator, it would be nice to be able to choose your preferred smoothing method (just as you can select the type of moving average: simple, exponential, or front weighted).
Just my 2 cents.
|
Worden Trainer
Joined: 10/7/2004
Posts: 65,138
|
StockFinder uses Wilder's Smoothing in its default Average True Range calculations and TC2000 version 12 uses Simple Smoothing in its default Average True Range calculations.
-Bruce
Personal Criteria Formulas
TC2000 Support Articles
|
Registered User
Joined: 7/30/2007
Posts: 1,072
|
Bruce, thanks for the clarification.
I've been using TC2000 more and StockFinder less lately, and I prefer Wilder's original Smoothing calculation
Do you have any PCFs lying around that calc ATR using Wilder's Smoothing? For example, it looks like the default TC2000 ATR has a period of 14 ... do you have a PCF for a 14-period ATR using Wilder's Smoothing? (Any period will do - I should be able to take your PCF and adapt it if I need a different period.)
|
Worden Trainer
Joined: 10/7/2004
Posts: 65,138
|
Please see the Please provide a description and PCF for calculating an ATR value to be used as a stop loss value topic for a formula. That said, you could just add a 1-Period ATR to the chart and add a 27-Period Exponential Moving Average to it (double the Wilder's Smoothing Period and subtract one to get the period of th Exponential Moving Average required to reproduce it).
-Bruce
Personal Criteria Formulas
TC2000 Support Articles
|
Registered User
Joined: 7/30/2007
Posts: 1,072
|
Thanks, Bruce. I spoke too soon earlier - I'm actually going to need a Wilder's Smoothed 14-period ATR for the next thing I'll be testing, so this link will be a huge help.
QUOTE (Bruce_L) That said, you could just add a 1-Period ATR to the chart and add a 27-Period Exponential Moving Average to it (double the Wilder's Smoothing Period and subtract one to get the period of th Exponential Moving Average required to reproduce it).
That's very interesting. I'll try it both ways to see how they compare.
|
Worden Trainer
Joined: 10/7/2004
Posts: 65,138
|
You're welcome.
-Bruce
Personal Criteria Formulas
TC2000 Support Articles
|