|
PureBytes Links
Trading Reference Links
|
Do you mean interesting in the profit sense or academic?
And what does one use for a measure of confidence? Would a test against
randomness suffice?
Pablo
On Fri, 21 Apr 2006 08:36:30 -0700, Alex Matulich <alex@xxxxxxxxxxxxxx>
wrote:
maybe i can question it better:
which day of the week has the highest probability to be a trend day
( or a
day with a greater range)?
i have read an article from larry williams where he wrote that days
after
monday, thursday or friday are trend days and any setup will have
greater
significance. so i thought how does he know that???
He knows that by writing software to identify setups and patterns
on different days of the week, and determine how often the setup
yielded fruit X number of days later. For example, suppose he
says "I wonder if 3 up closes in a row means anything?" So he
looks for places in the past 10 years or so of history that have 3
up closes in a row, record the day the pattern occurred, and note
what happened the next day and maybe 5 days later. In doing so, he
might discover, for instance, that if the 3rd up close occurs on a
Thursday, then Friday's price has a 72% chance of being lower.
Writing software for statistical analysis of patterns on market data
is drudgery but it can yield interesting results. In my opinion,
what's missing from all this is a measure of confidence (which is
not the same as probability).
-Alex
|