Wednesday, May 17, 2006

A Momentum Pattern That Seems to Work: The Relative Range

Here's an interesting application of the relative measurement concept that I've been toying with during the past several posts:

Suppose we look at the day's trading range as a function of the median daily trading range for the past 20 days. This relative range statistic tells us when a market move is very large or small compared to recent volatility norms.

I focused especially on large relative range days, because those also had the potential to be breakout days and perhaps lead to some momentum follow through. I also focused on the relatively low volatility market we've had since January, 2004 (N = 595 trading days); later explorations will look at a wider timeframe.

Since January, 2004, we've had 32 days in which the two day relative range in SPY exceeded 1.70 (i.e., today's range exceeded the median two day range by 70% or more). This was the case at the end of last week with the large drop. Two days later, SPY was down by an average -.10% (13 up, 19 down), which is weaker than the average two-day gain of .05% (320 up, 275 down).

Here is the interesting part, however. When the second day of the large two-day range was up from open to close (N = 11), the market was higher two days later on 10 of those occasions. When the second day of the large two-day range was down from open to close (N = 21), the market was down two days later on 18 of those occasions.

It thus appears that a large relative range leads to short-term price continuation in the direction of the large range move. I will need to explore this over a longer time frame and under different market conditions to determine whether this is a general or local pattern.


An observation from the Weblog: Energy stocks are actually below their levels from last January. Could it be that the fortunes of major oil companies will increasingly diverge from oil prices as countries nationalize their reserves?