|
PureBytes Links
Trading Reference Links
|
> has anyone experiences based on the logarithmic spiral / nautilus shell as
> described in the book
> "fibonacci applications and strategies for traders" by Robert Fischer?
Hi Wolfgang,
I spent some time trying implement this and found
several things. (I got some tips along the way from Ken
Churchill in the TradersGroup)
1. The formula in the Fischer book is wrong, or at least
it creates spirals that turn inwards, rather than outwards.
Outwards is what you want, because you want to specify
the centre and the radius and then spiral outwards.
In order to make the spiral turn outwards, the constant needs
to be set to 0.0765838. (This comes from the formula
natural_log(Phi)/(2*PI), 'where Phi = (sqrt(5)+1)/2 = 1.618034..)
2. You cannot implement the spirals in ELA code, for the
following reasons:
a. A single plot statement can only plot one point on any vertical
axis. I used the 4 plots to try to get more points displayed, but
it is close to useless.
b. The scaling cannot be set correctly. I got a spiral-ish shape,
but it was either "fat" or "thin" because of the different scaling
of the x and y axis. I believe it is not possible to set the
scaling correctly within an ELA.
3. I eventually implemented something in visual basic, and I
found the results were less than encouraging.
Spirals are heavily dependent on screen scaling. If you
try to place the same spiral on 2 different computers, the
results will not be the same if the physical screen
characteristics and/or the chart size are different.
This makes the results not repeatable. (let alone
backtestable).
4. I eventually decided that Fibonacci projections in
price and time needed to be kept separate to avoid the
scaling problems.
Bottom line - I found spirals pretty but useless in practice.
To see what I mean about scaling, imagine one of the
charts in the Fischer book is compressed in the y axis, and
unchanged in the x axis. (i.e. it is made "lower and fatter").
Applying the spiral to that chart will give a very different
result to that publised. To repeat the results shown in
the book, your screen will need to display the chart using
EXACTLY the same scaling as on the page.
Hope this saves you some time.
Peter
|