🚀 go-pugleaf

Hi Michaella,

I think it's necessary to differentiate between short and long term
prediction. Short term, e.g. about 2 - 4 trading days seems to be
predictable with probabilities of about 60 to 80 %. Long term is much lower.
Why short / long term? Mathematically there are a lot of influences of the
market, ending in daily quotes of relevant stocks. One of the attributes is
that the closer the time frame the higher the propability of a predicted
result. (variances) Viceversa: the longer the time horizone you're looking
for results, the higher the variances of all influences. Like a wave, that
is getting higher and higher, the range of possible results gets widened,
e.g. trustable results are more and more unpredictable.

Neuro methods use a network of artificial knodes. You have your data base
and then the neuro system tries to build a network of formulas around that.
That seems great, if that system tells you probabilities of 90% and higher.
But when you update your database, quite other formulas might be necessary
for high propabilities. But now the updated database-formulas are not
comparable to the earlier ones. Two databases -> two systems of prediction.
Each system is right, but not comparable for the next datas like quotes.
Neuro system try to overcome this problem by searching and using patterns.
If the stock quotes of xyz-company has a sinus-like wave each 15.th of a
month, then the system thinks it will forsee this wave. But due to the
variances it is very unsure that this works. The nearer the quotes you're
looking for, the better and more reliable the results.

Sum-up: don't use neuro system for long term trading. Being a day trader it
might work with propabilities of about 60 to 80%. But diversify to catch up
with the errors by variances. If you want to use a neuro system you should
try KnowledgeMiner. It's said the best system worlwide due to high number of
layers, and different parallel methods of prediction (fuzzy,
neuro-algorithms, system of equations).It is used by NASA, aircraft
companies and others for simulation and prediction. And its very simple to
use.

Regards
Thomas

"Michaella Po" <pelo@starmedia.com> schrieb im Newsbeitrag
news:pJ3c7.62664$oh1.24458481@news2.rdc2.tx.home.com...
> I'm an italian girl with a lot of questions about it. I found a club at,
> http://clubs.yahoo.com/clubs/neurostockinfoshare and all those inputs I
read
> there makes me feel so curious about stock prediction. I would like to
have
> some
> feedback. Any input will be nice and any information about neural
prediction
> will be helpful to me. I have downloaded the NS Shareware and some free
> databases trained from the links at the club and I think is good and easy
to
> use. Any comments about?
> Michaella....
>
>
>