John Kevin -
Developer of the market’s first Talking S&P robotic market analyst “Cybornia”, and
the markets first Multi-
All Rights Reserved © Trader Robotics
Contact: info@traderrobotics (dot com) Trader Robotics provides financial market forecasting Videos and is based in Missouri USA.
After reading a good article on another website “System Trader Success” by Jeff Swanson (no affiliations) Entitled “What’s wrong with ATR Stops? Please read his article before this one.
I decided to do a quick write up myself on the ATR subject.
My results found good reason to consider using ATR related stops. Jeff’s not so much, though all of his points were absolutely true and valid.
My explorations of the ATR stop produced 3 additional concepts that you may find interesting.
Jeff’s ATR(100) indicator used in his article study would be too short of an average for most benefits.
After deeper digging, the benefit I found in using the ATR stop is on a longer term basis.
Different data sets will have different ATR Means. This is where the ATR comes in useful. The advantage of the ATR related Stop, is to identify and tap into, the volatility on a trending basis. Volatility trends will be more accurately seen on longer ATR averages. I found the alternative use of the ATR would be NOT to look for sudden changes in a shorter ATR such as 100 bars,, but rather on a longer term trend of the ATR. Doing this adjust’s your strategy on a more gradual scale to better fit a particular data set’s volatility conditions.
OOS (Out of sample) results often change drastically from back testing results, due to a change in longer term volatility. A longer term ATR will help automatically adjust your strategy to the varying profiles of different data sets.
Concept 1. An ATR bar average could be more helpful as a 1,000 or 2,000 ATR average. 200 would be the bare minimum. My testing results were improved across most data sets by implementing a several thousand bar ATR average.
Concept 2. Another idea is to have the averages for different periods of the day.
Such as an ATR(24) applied only during 9 am to 3 pm, which is then added to several
previous days ATR averages for that same 9-
Concept 3. In addition, a combined Fixed dollar stop in combination with an ATR weight, would be another choice. Such as a $250 fixed dollar + a smaller multiple of an ATR(), added to it. This concept reduces the impact of the ATR without foregoing it altogether.
Jeff pointed out the variance of range in the 100 bar ATR was too great to be beneficial. I agree. However it’s that exact fluctuation of the ATR’s range that offers the greatest benefit. The only problem was that on Jeff’s test, the ATR(100) was too short of a time frame. So, the range of the ATR was too great in relation to the time frame (length of the average) The longer we extend we the ATR out, the less volatile the range will be. There is a Time/Price relation within the ATR, so adjusting one axis (horizontal axis = Time) affects the covariance of the average.
P.S. The same concepts can sometimes test effective in adjusting entries, and look back periods for other indicators. Use of ATR on longer averages usually improves my testing results where ever I apply it to.
Using ATR to improve trading models.