## ATR Value

Discussions about the testing and simulation of mechanical trading systems using historical data and other methods. Trading Blox Customers should post Trading Blox specific questions in the Customer Support forum.
Forum Mgmnt
Roundtable Knight
Posts: 1842
Joined: Tue Apr 15, 2003 11:02 am
Contact:
Menelik wrote:Let’s try your example, to see if what you say is true.

Percent Account for N = 1%
Entry Breakout = 55
Exit Breakout = 20
Stop in N = 2.0

100,000 portfolio. 1% risk, or \$1000.

A) \$25tradable. \$2.48atr N = 2.0
B) \$82tradable. \$4.80atr N = 8.0

stock A.
\$25 stock. @ 2.48atr = \$5 stop. 200 shares to risk or \$1000.

Stock B.
\$82 stock @ 4.80atr =\$38 stop. 26 shares to risk or \$1000

A) N factor of .10
B) N factor of .06

(A) is thus riskier than (B)!
Howard was correct. The sizing in shares would only relate to the ATR (remember N is the same concept as ATR). So with a 1% for N sizing you'd have:

\$1,000 / \$2.48 = 403 shares of Stock A, NOT 200
\$1,000 / \$4.80 = 208 shares of Stock B, NOT 26

So you'll have:

\$25 X 403 = \$10,075 invested in Stock A
\$82 X 208 = \$17,056 invested in Stock B

You can't literally have:

\$2.48atr N = 2.0 - If ATR is \$2.48 then N is \$2.48, they are the same.
\$4.80atr N = 8.0 - If ATR is \$4.80 then N is \$4.80, they are also the same.

Note: The Turtle System would not have used different sized stops for different stocks, they would have used the same stop across all the stocks. The example Howard gave was a 2N stop, so the stop for both would be 2N or 2 ATR. Your example with a 2N stop for Stock A will represent about 2% of the account. An 8N stop for Stock B will represent about 8% of the account

What the Turtle System does is use N (or ATR) as a proxy for volatility and risk. Positions are then sized to equalize that risk across each position for every market. One could argue that ATR is not a good measure to use to represent risk, that other measures better represent risk. While that may be true, it does not mean that the Turtle System was not a valid and serious attempt to equalize risk across markets.

N also works reasonably well as a way of sizing the stop. As you have pointed out, a low priced stock with relatively high ATR to price ratio is relatively more volatile. Using ATR-based stops would tend to account for this increased volatility and allow larger stops for the stock than a pure percentage based stop.

There are certainly other ways to account for the volatility, and other ways to account for risk. However, the Turtle approach was very advanced considering we are talking about an approach that was devised more than 20 years ago.

Kiwi
Roundtable Knight
Posts: 513
Joined: Wed Apr 16, 2003 1:18 am
Location: Nowhere near
I have to admit to feeling a bit of intellectual insufficiency here!

I thought I was a bit of an intellectual but I've realised that I might just be
a boring pragmatist. To me this feels like we are trying to extract too
much meaning from the real data available.

I think that you can make all sorts of interesting arguments about the true
risk of a position but in the end they do not matter! To me there are two
levels of risk I need to consider:
1) My normal trading risk - this is the distance to my stop + a bit of
slippage whether I set it by atr, 10 day high, dollar amount or whatever.
2) Abnormal event risk - what are the effects of a nuclear bomb, an oil
shock, etc and is it likely that they will equally favor and disadvantage me
or will it all go against me.

So I don't use anything too clever to decide my risk: "robust statistics"
only please. I trust in diversification to make my normal trading risk a
good estimate of what I will experience in real time. I add some thought
about nuclear events to this and as a result try to position myself so that I
will gain on some system types and lose on others. I keep my eggs in
separate baskets so that even if I loose everything on my long term
trading my other systems and cash may be available and functional. And
I can start again.

After a long and interesting thread: Just my humble opinion ...

John

MCT
Roundtable Knight
Posts: 102
Joined: Fri May 16, 2003 7:27 pm
Forum Mgmnt,

The thread is about ATR-Value & exposure and the discussion resulted from a Post dated: Fri Sep 12, 2003 7:12 by Nathan:
100,000 portfolio. 2% risk, or \$2000. 2atr stop

stock A.
\$50 stock. 5atr = \$10 stop. 200 shares to risk \$2000.
dollar commitment on trade = 50 * 200 = \$10,000.
Max theoretical loss = \$10,000

Stock B.
\$10 stock 5atr = \$10 stop. 200 shares to risk \$2000
dollar commitment to trade = 10 * 200 = \$2000
Max theoretical loss = \$2000

Menelik in response wrote:
Your position size: = 2% or \$2,000/ \$10stop (\$5atr x 2atr) = 200shares

What Iâ€™m saying is: \$2,000/\$5atr=400 shares

You were comparing to different systems in your post.

Iâ€™m not an expert of the turtle system. I was simply referring to Howard. Given the above, Howards reasoning would have resulted in my response you were refering to in the previous post.
Howard said:
It just so happens that "N" is used both to size positions, and to place stops. Neither has an impact on how positions are sized. Stop in N can be changed from 2.0 to 4.0, or even 8.0, with no impact on position sizing.
If I have misunderstood Nathanâ€™s original example â€¦ my apologies to him.
However, the Turtle approach was very advanced considering we are talking about an approach that was devised more than 20 years ago.
After having studied the system for the past two weeks â€¦ I have to say I couldnâ€™t agree with you more. I doubt if there are other systems that come close when it comes to equalizing exposure â€¦ I take my hats off to Richard D. and William E. Itâ€™s truly astonishing they were trading such systems 20 years ago. Having said all of that, it is just my personal opinion that strictly mechanical systems will always face uneven risk exposure. I donâ€™t believe there is a way around that. But everyone ought to perform their own tests â€¦ to arrive at a reasonable conclusion for themselves.

Simply put, itâ€™s my personal opinion that systematic stops & exits in general are more likely to make antitrend decisions. At least that has been my experience.

Question for c.f.:
Which exits were hit more often, N exits or breakout exits, when you were trading as a turtle?

Best regards,

MT

p.s And one last point, my comments in reference to you were strictly within the context of the atr concept. For example, atr or N does not provide any meaningful information "until" it has been considered relative to price. That's why I said earlier in the thread that for any given instrument it has to be looked at atr/close, I believe that provides more meaningful information.In your case or the turtles, it seems the volatility units were normalized by multiplying the shares by the price of the stock, which is essentially the same thing. What I'm saying is
without the "price" to be paid, atr or volatility does not mean very much, thus it is only meaningful "relative" to price. My point all along has been it has to be looked at relative to price. I feel such information is quite easy to over look ... thus this thread.

Nathan
Roundtable Knight
Posts: 127
Joined: Sun Jul 13, 2003 3:52 am

### atr

1. What I said on the 12th, or the point I was trying to make, has not really been discussed since.

2. Menelik, I have been using 1atr, or as the turtles call it N, to calculate position size in my examples. That is what stays the same and is why both howard and I have said position size does not depend on the stop.

On the 19th, i said this:

"The model has attempted to equalize the volatility impact on the portfolio through Position sizing, not stop placement"

ON the 20th, howard said this:

"It just so happens that "N" is used both to size positions, and to place stops. Neither has an impact on how positions are sized. Stop in N can be changed from 2.0 to 4.0, or even 8.0, with no impact on position sizing. The two are separate."

ON the 21st, c.f. said this:

"Positions are then sized to equalize that risk across each position for every market."

It seems as though the same point, or approximately the same point, has been made by several people. The impact on the portfolio is equalized by position sizing using atr, or as the turtles call it N.

This may have not been clear in my examples because i skipped a step, using 2% without first calculating the N value independantly. However, In the end it makes no difference as it is mathematically the same thing. Using the latest example:

\$1,000 / \$2.48 = 403 shares of Stock A

or, skipping a step using the 2% risk parameters of my example:

2000 / 4.96 = 403

or, an account that risked 3%

3000 / 7.44 = 403.

as can be seen, the position size is not effected.

Nathan

Howard Brazzil
Roundtable Fellow
Posts: 54
Joined: Wed Apr 16, 2003 12:45 pm
Location: Houston, TX USA
Hey guys,

I've been miquoted twice now, and it makes
what I wrote sound like a contradiction:
MISQUOTE: It just so happens that "N" is used both to size positions, and to place stops. Neither has an impact on how positions are sized.
Stop in N can be changed from 2.0 to 4.0, or even 8.0, with no impact on position sizing.
Here's what I actually wrote. The critical missing sentence
(in blue) has been restored. As you can see, it changes the
meaning of the passage dramatically:
Actual Quote: It just so happens that "N" is used both to size positions, and to place stops...

The actual risk on the trade is (approximately) defined by the distance from the entry price to the Stop in N, or the
exit breakout, whichever is closest to price.

Neither has an impact on how positions are sized.

Stop in N can be changed from 2.0 to 4.0, or even 8.0, with no impact on position sizing. The two are separate.
Big difference!

Nathan
Roundtable Knight
Posts: 127
Joined: Sun Jul 13, 2003 3:52 am

### mistake

Sorry about that howard, I made the lasy and very stupid mistake of not looking up the quote from the original source.

Nathan.

MCT
Roundtable Knight
Posts: 102
Joined: Fri May 16, 2003 7:27 pm
Nathan:

100,000 portfolio. 2% risk, or \$2000. 2atr stop

stock A.
\$50 stock. 5atr = \$10 stop. 200 shares to risk \$2000.
dollar commitment on trade = 50 * 200 = \$10,000.
Max theoretical loss = \$10,000

Stock B.
\$10 stock 5atr = \$10 stop. 200 shares to risk \$2000
dollar commitment to trade = 10 * 200 = \$2000
Max theoretical loss = \$2000

Menelik said:
My earlier response to you was:
Your position size: = 2% or \$2,000/ \$10stop (\$5atr x 2atr) = 200shares

What Iâ€™m saying is: \$2,000/\$5atr=400 shares
Obviously there is a major misunderstanding between us â€¦
my apologies to that. And that is a result of â€œdiscussingâ€

NJP
Contributing Member
Posts: 7
Joined: Thu Apr 17, 2003 3:13 pm
Location: Florida
ATR tells us one thing, the volatility of the market over the last 7 or 10 or 22 or whatever number of days we want. Volatility is not a bad thing. We need volatility to be present to make a profit. I use ATR to tell me "How much?" and that is it. A \$10 stock with a \$5 ATR is no worse of than a \$50 stock with a \$5 ATR. If the \$10 dollar stock was at \$7 one week ago and the \$50 stock was at \$48 then I would rather take the trade on the \$10 stock. A \$50 stock with an \$5 ATR could have been trading between \$47 - \$52 for the last year or it could have been \$35 a year ago. ATR does not tell us that it simply says, How Much.

Forum Mgmnt
Roundtable Knight
Posts: 1842
Joined: Tue Apr 15, 2003 11:02 am
Contact:
Menelik,

One of the reasons I really like mechanical trading systems is that I can ask "what if" questions and get an answer from the markets. It may be an incomplete, flawed, and caveat filled answer, but it is at least an answer.

I can program different sizing algorithms and see which ones deliver the best risk/adjusted returns and I can do this using my own measure of "risk/adjusted returns".

I see some merit in looking at other measures of risk besides ATR or standard deviation based short-term volatility.

Have you done any historical testing which would indicate that an ATR based sizing approach does not adequately represent the risk of a market and that there are significantly better approaches?

It will probably be a few more months before this kind of research hits the top of my pile, as for the moment I am finding other research more fruitful.

- Forum Mgmnt

NJP
Contributing Member
Posts: 7
Joined: Thu Apr 17, 2003 3:13 pm
Location: Florida
MT,

You said,
"But what Iâ€™ve been saying is 1) the risk in each one is different and they should either be sized differently
"

Why? I agree with what you wrote further down.
"Risk of loss is proportional to ATR"
Also, you state
"speed of price increases as ATR decreases"

I have to disagree with that as I have found ATR and price to be independant. Usually, but not always as the market chops sideways toward the end of a good move the ATR is considerably higher than during the move.

NJP

bagherra
Senior Member
Posts: 33
Joined: Fri Apr 25, 2003 9:03 am
Location: Charlotte, NC

### Exotic Power-Law?

Menelik,

What is an 'exotic power-law distribution'? It sounds cool.

Back testing has serious drawbacks, as the discussion of the deterioration of PGO's results over the past ten years demonstrates. There is a danger of taking too large a position size on the basis of backtested results and blowing through your backtested maximum drawdown. However, I think that the chief benefit of backtesting may be the development of one's intuition. All that work is the equivalent of getting your hands dirty with the numbers, and may lead to beneficial discretionary decisions.

MCT
Roundtable Knight
Posts: 102
Joined: Fri May 16, 2003 7:27 pm
"speed of price increases as ATR decreases"
I have to disagree with that as I have found ATR and price to be independant. Usually, but not always as the market chops sideways toward the end of a good move the ATR is considerably higher than during the move.
NJP,

Your answer is a simple inversion of what I said; the relationship holds. I believe, the speed of price is derived on a causal, deterministic basis. The uncertainty ATR measures is â€œautomaticallyâ€

Dan G
MT - Your posts are very difficult for me to understand.

MCT
Roundtable Knight
Posts: 102
Joined: Fri May 16, 2003 7:27 pm
mickslam,

Let me begin by saying ... sorry. My posts focuse on the philosophical and theoretical aspect of things...I find that more important than simple number crunching. I believe, we have to theorize and philosophize in order to extract real meaning from our sample tests...

Many serious traders have thrown up their hands and declared market phenomena to be inexplicable. They simply go with the equations because they â€˜workâ€™. They abandon any effort to understand why the equations work or to develop a physical picture of a system or market phenomena. Such examinations are what allowed man to land on the moon or build an atom bomb. And no, Iâ€™m not referring to fundamental analysis. To the contrary, Iâ€™m referring to a mental process of simplifying which allows one to build highly 'robust' and simple methods such as the turtle system. Reality has to first be understood and imagined ... only then can we build robust models that reflect it. Just my \$0.02

Best Wishes,
MT

P.s I had the feeling that maybe this wasn't the right forum to present such ideas-nonetheless, I elected to present them.
Last edited by MCT on Sun Oct 12, 2003 11:38 pm, edited 1 time in total.

Kiwi
Roundtable Knight
Posts: 513
Joined: Wed Apr 16, 2003 1:18 am
Location: Nowhere near
Menelik,

Keep presenting them. They are interesting and add a different persepective even if, at times, I find them inexplicable I frequently found that with engineering reports from one of the brightest guys I ever met but it didnt stop me working with him.

Mick, I liked your idea on MAE but didnt think I could add anything at present. Menelik, what is your view on that area?

John

MCT
Roundtable Knight
Posts: 102
Joined: Fri May 16, 2003 7:27 pm
Hi John,

There are few discretionary traders on this forum and it feels odd at times ... the irony is, there is more to learn from mechanical trend followers than most other types of traders â€¦

What Iâ€™m suggesting is quite radical. Many such as William Eckahardt and Nassim Taleb have suggested the human mind is incapable of understanding randomness and searches for patterns where none exist. To quote W.E in The New Market Wizards:
Do you use chart patterns in your systems?

Most things that look good on a chart-say, 98 percent-donâ€™t work.
Why is that?

The human mind was made to create patterns. It will see patterns in random data. A turn of the century statistics book put it this way: â€œtoo fine an eye for pattern will find it anywhere.â€

trackstar
Roundtable Fellow
Posts: 76
Joined: Sun Mar 09, 2008 11:03 pm
Location: Houston Texas
Are any of the participants of this thread still interested in taking this discussion further? I know I am very late to the party. I am attaching a picture of my attempt to replicate for futures what Menelik described as %ATR(yellow) and the Variance of %ATR(blue) on a daily chart of comex Gold.

I can also post the code in EasyLanguage if anyone wants it.

edit: A little clarification on how I decided to apply to futures. ATR measures volatility, and that what Menelik suggested was that it needed to be graded against the price. I took the ATR and divided it by the total number of ticks that would be passed through if the commodity went to 0. By no means am I inferring that I have a good method of replication here. My goal was to keep it simple and structured in a way that could be applied to all commodities.
Attachments
Gold.JPG (132.67 KiB) Viewed 5783 times