|
PureBytes Links
Trading Reference Links
|
Mark,
>I got interested in the question "which futures markets are the
>trendiest?" I calculated Wilder's "ADX" indicator (which measures
>trend strength) for each day since Jan 1, 1980, and found the
>average daily value of ADX for each market. The idea behind this
>calculation is: the trendiest markets will have, on average, the
>highest values of ADX. So I ranked markets according to their
>average value of ADX.
Well, I guess that's one way to answer the question.
I've usually seen the question answered through measurements of
statistical variance and comparing to the expectation from a
random walk. A pure random walk, whether constructed from uniform
or gaussian-distributed random numbers, will always exhibit the
characteristic that standard deviation scales (increases) in
proportion to the square root of time.
This implies that one can measure whether the trends in a market
are stronger or weaker than the trends which appear in a purely
random-walk market. The idea is, the "trends" in a random market
are meaningless and untradeable, but something trending more
strongly might be taken advantage of.
>There are a few surprises.
>
>Japanese Yen futures don't appear in the top ten. Yet many people
>consider JY to be the single trendiest market of them all.
What happens when you apply ADX to a random walk?
Regardless of what the ADX ranking suggests, JY happens to have
(or did have, up to the early 1990s when I read the analysis) the
largest deviation from a pure random-walk. That is, the trends in
JY are generally stronger than those exhibited by a random walk
having the same variance.
This is described in "Futures According To Trend Tendency" by
E. Michael Poulos, TASC January 1992, p. 61. (TASC was a good
magazine then.)
Poulus examined several different lookback lengths and calculated
"channel height ratio" using the following procedure. For an
N-day channel length, start at Day N and record the maximum range
(highest high minus lowest low) during that interval. Repeat for
Day 2 through Day N+1, 3 through N+2 and so on. Average all these
heights. Calculate the channel height ratio is obtained by dividing
the average N-day channel height by the average 1-day channel
height. Repeat for other lookback lengths.
The channel height ratio for a pure random walk is simply the square
root of N.
Comparing the channel height ratios for various futures contracts
against a random walk, one finds (at least up to 1992) that Yen
consistently exhibited the highest channel height ratio. Wheat had
the lowest, although wheat still exceeded a random walk for short
lengths, but not long lengths.
I have been meaning to repeat this test and see what has changed
in the last 12 years. Who knows, perhaps it might agree with your
assessment using ADX.
--
,|___ Alex Matulich -- alex@xxxxxxxxxxxxxx
// +__> Director of Research and Development
// \
// __) Unicorn Research Corporation -- http://unicorn.us.com
|