I have been studying the market with different statistical functions for many
years ; binomial law, standard deviation, normal law, etc. I've also
downloaded many programs trying to understand the behavior of the market and
it's not easy, just predicting the direction of the market is a task even if
at the end the risk/reward could generate lot's of profits or losses; how
many people brokers, individuals had predicted the black Monday in 1987 ?
Anyways why a 6th degree, why not 5th or 7th degree ? I believe you would
extract a minimum or a maximum like in a second degree equation (-b/2a for x)
if negative maximum (downslide) and if positive (uptrend) so this would give
i believe a range over time ????? i do hope i'm making sense here......!!!!
this is order normal conditions but what if in 5 days the barrel of oil would
go up or down by $3.00, what if the $us would trade at 115 yens, if interest
rates would increase/decrease of 1%, if Microsoft or whatever stock on the
Dow would increase/decrease his earnings of $1 how would the market react ????
So far one of the best thing i came over was a multiple regression (LINEST
function) not only you can play with the amount of periods, variables but
also you can play with the standard deviation of the equation and better yet,
you can see what impact each variable does in your equation. Believe me
currencies have a lot of impact for example the price of oil is in $US
however when converting into foreign currencies that price could be
attractive/expensive, same thing for gold, interest rates, options on
indexes, etc.
I remember in college doing a paper on a simple regression unemployment rate
and gallons of beer sold, it was at 0.9 over a long period period of term
(more then 30 months) it was logical that people since they dont work stayed
home and drank beer however people had lower income !!!! now if you would
join a third variable the correlation would increase or decrease a lot,
however if a fourth variable would be had up the correlation would stabilise
around 0.6. Even if you would change the 3rd or the 4th one, it would stay
around 0.6.
it seems also that you have a hard time with small numbers probably because
when using the date, Excel uses a number instead; 39000 represents Oct
10/2006, use Oct 10/1906 and it will recalculate it at 2475, so small numbers
will come lot less and reformat it just to see the date without the year.
Also dont forget to use the format scientific because if you have a variable
like 1.5666 E-25 it takes a mighty big number just to came to 0,01 (power
22-23-24).
Since June 2005, by using multiple regressions i have a compounded annual
rate of return of 16.05% wich would place me in Canada among in the first
best quartile of managers even if my worksheet uses 60 megs and i believe if
i fine tuned variables, periods, ANOVA, decision trees, study sectors, have
the time also to compare forecasts and results i would out perform the
general trend of the market.
On the uptrend it's easy to buy stocks and out perform the index but it
comes harder and harder to say; time to take to my profits and run away and
stay away, on the side line until a stock reaches a buttom and buy it back
when it reaches the buttom unless you're into put options.
last week i have sold some units of a Japaneese fund that tracks the Nikkei
index that i had bought in October 2006 at $5.5 and sold them at $6.80
because the probability of reaching well over $7 was decreasing but on the
other hand the probability of increases in the price of gold, oil, bonds,
high-tech, small caps was slowly increasing even if the market was high so
with the income i have reinvested 50% into a fund that tracks the canadian
index and kept 50% on the sideline in case if.........