Chuck LeBeau's System Traders Club
BULLETIN Vol. 1 Number 10 Oct. 20, 1998
Table of Contents:
This will be the first in a series of articles about using Average True Range. I will be speaking on this topic at the 20th Annual TAG conference in Las Vegas November 21st, 22nd, and 23rd. I thought our Traders Club members might enjoy these excerpts from my lecture notes. For information about the TAG conference (highly recommended) visit their web page at http://www.telerateseminars.com
Average True Range
Average True Range is an indespensable tool for designers of good trading systems. It is truly a workhorse among technical indicators. Every systems trader should be familiar with ATR and its many useful functions. It has numerous applications including use in setups, entries, stops and profit taking. It is even a valuable aid in money management.
The following is a brief explanation of how ATR is calculated and a few simple examples of the many ways that ATR can be used to design profitable trading systems.
How to calculate Average True Range (ATR).
Range: This is simply the difference between the high point and the low point of any bar.
True Range: This is the GREATEST of the following:
1. The distance from today's high to today's low
2. The distance from yesterday's close to today's high, or
3. The distance from yesterday's close to today's low
True range is different from range whenever there is a gap in prices from one bar to the next.
Average True Range is simply the true range averaged over a number of bars of data.
To make ATR adaptive to recent changes in volatility, use a short average (2 to 10 bars). To make the ATR reflective of "normal" volatility use 20 to 50 bars or more.
Characteristics and benefits of ATR.
ATR is a truly adaptive and universal measure of market price movement.
Here is an example that might help illustrate the importance of these characteristics:
If we were to measure the average price movement of Corn over a two day period and express this in dollars it might be a figure of about $500.00. If we were to measure the average price movement of a Yen contract it would probably be about $2,000 or more. If we were building a system where we wanted to use the set appropriate stop losses in Corn and Yen we would be looking at two very different stop levels because of the difference in the volatility (in dollars). We might want to use a $750 stop loss in Corn and a $3,000 stop loss in Yen. If we were building one system that would be applied identically to both of these markets it would be very difficult to have one stop expressed in dollars that would be applicable to both markets. The $750 Corn stop would be too close when trading Yen and the $3,000 Yen stop would be too far away when trading Corn.
However, let's assume that, using the information in the example above, the ATR of Corn over a two day period is $500 and the ATR of Yen over the same period is $2,000. If we were to use a stop expressed as 1.5 ATRs we could use the same formula for both markets. The Corn stop would be $750 and the Yen stop would be $3,000.
Now lets assume that the market conditions change so that Corn becomes extremely volatile and moves $1,000 over a two day period and Yen gets very quiet and now moves only $1,000 over a two day period. If we were still using our stops as originally expressed in dollars we would still have a $750 stop in Corn (much too close now) and a $3,000 stop in Yen (much too far away now). However, our stop expressed in units of ATR would adapt to the changes and our new ATR stops of 1.5 ATRs would automatically change our stops to $1500 for Corn and $1500 for Yen. The ATR stops would automatically adjust to the changes in the market without any change in the original formula. Our new stop is 1.5 ATRs the same as always.
The value of having ATR as a universal and adaptive measure of market volatility can not be overstated. ATR is an invaluable tool in building systems that are robust (this means they are likely to work in the future) and that can be applied to many markets without modification. Using ATR you might be able to build a system for Corn that might actually work in Yen without the slightest modification. But perhaps more importantly, you can build a system using ATR that works well in Corn over your historical data and that is also likely to work just as well in the future even if the nature of the Corn data changes dramatically.
More to come in our next Bulletin
MORE MEMBER FEEDBACK ON VOLATILITY
(See special Bulletin #8)
The Babcock Exit
I think it's great that you and some subscribers are responding to this problem. Although I don't know how to program the following trailing stop I thought it is worthy of your consideration. The source of the stop methodology is Bruce Babcock from whom I have purchased a couple of systems.
Basically his approach to a trailing stop would be to set a percentage of profit that one would like to retain after a certain level of profits is reached. For example, in the bonds you would set it up to retain 25% for every $1000 or maybe $1500 worth of closed profit.
The code would be written such that each person could set their own risk parameters by adjusting the retention percentage and/or the dollar increment level. I find this approach more comfortable to my trading logic than using an average true range approach because Bruce's method will work even if the market retraces without an abnormal increase involatility.
A Timely Warning from Rick Saidenberg
In part of Richard Sheeler's note, he suggests optimizing with the Trade Station % Trailing Stop.
<< A more significant improvement results from just adding the Trade Station % Trailing Stop to the systems as follows:
Big Dipper 10% with $6,000 floor.
25X25 15% with $5,000 floor (very good!).
Serendipity 15% with $1,500 floor.
7-11 Yen 15% with $6,000 floor.
BEWARE! This stop leads to historical results which will not match the trades you will get while following the system real-time. This situation occurs when a LongExit occurs on the same bar when the high satisfies the floor; or when a ShortExit occurs on the same bar when the low satisfies the floor.
Response from Chuck
<< Is this the infamous "bouncing tick" problem I hear so much about? >>
No. This is different from Bouncing ticks.
% Trail Stop In historical testing, for a long exit, TS uses the high of a bar to calculate the trail stop exit level. It then reports the exit price on that same bar calculated from the high. In real-time, for a long exit, as soon as the Floor level is reached, TS places a stop to exit the trade - this is before the completion of the current bar (In Tools-Options-System, you can set how frequently the trail stop is recalculated). If the market turns down enough to hit the trail stop, TS stops you out of the trade correctly. However, if the market subsequently turns up to make a new high for the bar, the historical test will show an exit at a higher price.
TS uses only the OHLC data for historical testing. So whenever there is more than one trade in a bar (entry and exit count separately), since TS does not resolve to the tick, TS is prone to giving historical results which do not match the results in real-time. Bouncing Ticks is for historical testing. It is an attempt to guess the sequence of ticks in one bar. It only makes an assumption, so it is never precise. In order to be confident in a historical test, there must be a maximum of one trade in each bar. This may require writing code so a system can operate on a small enough timeframe so each bar contains a maximum of one trade.
New System on the Horizon
We are almost finished working on a Bond system designed to trade in sideways markets. The "Sidewinder" Bond System is designed to complement our existing bundle of Bond Systems. We expect to announce the release of this new system very soon. Our research to date looks very encouraging.
That's all for now. Many thanks to all those that contributed ideas and
comments. Your help makes my job much easier.
Good luck and good trading