The issue I have is predominantly my inability to access the capabilities
of Amibroker.What i am realising is the other programs I have used were
toys compared to Ami,but I could access most of what the programs had to
offer.Most of the functions were hard coded to make it easy.....
Ami can be simple if you stick to basic AFL,or exceedingly complex if you
want to raise the bar.The ATC index calcultion is a good example.You read the
users manual,you look at the ATC doc and you think you have a clue.But the
truth is,you are only guided down a path..You run into complexities you never
knew existed because there are so many things Ami can do.Tomasz statement
of "when are users going to realize there is nothing Ami cant do",is not
a boast.Its a simple fact.Unfortunately,for the new user/programmer this can
be quite the challenge and most time/energy consuming...
I dont think you fully understand my issue with ATC and Indicies.You need
to create an ATC composite on an industry with very few stocks.i.e
<6.What you will see,is if there are stocks with varying data history,if
one of the shorter history stocks have an initial price that is substantially
different than the index you will get a large spike.It has nothing to do
with:
It is simply how a price weighted index is calculated.Run thru your
database and take a good look at your industry groups with different data
histories.IMHO,pad and align is not the answer.What is the answer is for me to
learn how to calculate a different type of index
To be specific,I am am ex user of AIQ.IMHO,they were the "pioneer" in
Sector/group/stock data formats and made everything fairly
seemless.They had one method of creating and index.To the best of my
memory,it was equal weighted,the index would have an initial value of 100 and
changes were the average of the daily percent returns.I am not saying this
methodology is right or wrong,but you will NOT run into the possible dilemna
of massive spikes in your index as you do with a price weighted index.
----- Original Message -----
From: "brian.z123"
<[EMAIL PROTECTED]>
Date: Saturday, September 16, 2006 4:46
am
Subject: [amibroker] Re: ATC mis-Calculation??????
To:
[email protected]> Allan,
>
> No you are not
nuts!
>
> How can an *index* include a stock that wasn't trading
back then?
> It is how you choose to handle these exceptions that
decides
> what
> kind of trader you are.
> Howard Bandy's
example chapter, from his Quantitative Trading
> book,
> gives
one example of how this problem can be managed when
>
backtesting.
> There are others.
> Assuming you adopt that
method,will it be accurate?
> Will it be more or less accurate than the
headline indexes?
>
> What do you want the index for
anyway?
> Backtesting or scanning.
> In real time the indexes are
always *correct* unless a company
> is
> suspended.
> In
that case, as Tomasz pointed out in a previous post, the
> index
> pads the data for the suspended symbol.
>
> If you buy
historical records of an index, who was keeping track
> of
> the
constituents at the time, and can you obtain that info?
> If a company
is in an index and the nature of it's business
> changes
> who
decides if and when to remove it from the index?
>
>
BrianB2.
>
>
> --- In
[email protected],
"matrix10014" <[EMAIL PROTECTED]>wrote:
> >
> > Hello all,
>
>
> > I have received various thoughts on ATC and spikes(pad and
> > align,reload data),but IMHO i think their may be an inherent
> flaw
> in
> > the calculation of indicies..There is
also a very good chance
> i am
> > losing it...
> >
> > Here is my real issue regarding ATC...After running the simple
> ATC
> > code,i found a tremendous number of data spikes.It
appears to
> be
> > occuring when a stock with a shorter data
history is in the
> > composite.If I am not mistaken,ATC sums up the
indivisual
> stocks
> and
> > creates a price weighted
index.So,if we have a an index of
> four
> > stocks,A,B and C
having one year of history,and stock D having
> 6
> >
months,if stock D's initial price differs substantially from
> the
> > composite of A,B,C,there will be a large spike.I verified
> it.As an
> > example..On day 180,A+B+C=50.On day 181,if
stock D is
> introduced
> and
> > it is trading at
50,A+B+C+D =100 and the index will have
> increased
> >
100%.Of course if you include the count divisor,this affect
> will
> be
> > mitigated,but none the less,it is not the proper way
to handle
> an
> > index.IMHO, an equal weighted composite
with the change in
> percent
> > returns may be the
solution
> >
> > Ami I nuts,or just losing my
mind???
> >
> > Allan
> >
>
>
>
>
>
