ATR Testing

We've covered some of this before but I like to revisit favorite themes every now and again to ensure that new readers have a chance to understand what we do here.

Today I'll discuss the advantage of using the average true range which I use as a 10-period averaged value or, ATR(10). I have reasons for using the 10-period - first it is the one most commonly defaulted on the various software packages I access but second - I've tested other lengths and the 10 always comes up the winner.

When I'm day trading from the minute charts and I have a couple of different selections I always like to use the one(s) with the highest ATR(10). However, when swing trading and trying to select between stocks I like to go with the lowest ATR(10).

The ATR is an excellent proxy for volatility and if you are looking for a quick 50 cent jump you want high volatility. On the other hand if you want a sustained rise over several days you want to start from a low volatility position.

This can be proven and I wrote a simple test to do so. I tested the 4 day to 30 day returns of stocks trading between 15 and 35 dollars with a moderate 90-day average volume of 500000 shares for each of the two following conditions -

ATR(10) reached a new 26-week high = 49% win percentage, .63 Reward/Risk, and -62% ROI;
ATR(10) reached a new 26-week low = 62% win percentage, 1.52 R/R, and 49.74% ROI.

While the low model achieved a much better return than the high model it really isn't anything to cheer about.

But I did capture this low method in the previous filter post - A New Way to Look at RSI(2) and ATR(10). And while I'm not using a 26-week low I am using a low reading that is unbounded.