|
PureBytes Links
Trading Reference Links
|
Conrad --
I agree it appears that ratio data would change the height of bars
due to inflation over the years, if nothing else. Today's bar in
terms of height would be "bigger" in comparison to other methods, I
guess, but it seems to me only in terms of what you see on a chart.
Otherwise, it represents exactly the same relative move as the others
do. That's the idea, I gather. If a price move today would represent
2%, then the idea is to have past data representing a 2% move also
show as the same 2%. Therefore, the argument as I understand it is
that this is a more accurate way to look at a market.
It also is supposed to more accurately depict the P/L in that it has
changed dollars and cents to percentages that will translate into
apples-to-apples dollars regardless of when you're looking at for
testing. However, your point about points or $ stops being a problem
sounds like a good one.
But just as you said, I'm no expert either, which is why I have
posted my questions. I'm certainly having some trouble visualizing
how all of this would work in comparison to other methods. That means
I don't know how to react to your other suggestions, so I'm still
hoping for more info.
Also, it's interesting you mention CSI. I've heard good things about
them, so I contacted them with my questions as well. I was sent some
e-mail articles that I'm still digesting. They seem to be
particularly high on this ratio (proportional) data. If it would be
appropriate, I could post it here. Would CSI object? Should I do it?
(I have no connection with CSI or any other vendor of any kind.)
Jim
----Original Message Follows----
From: cb <cpbow@xxxxxxxxxxxxx>
Reply-To: cpbow@xxxxxxxxxxxxx
To: jdjdjd@xxxxxxxxxxx
CC: RealTraders Discussion Group <realtraders@xxxxxxxxxxxxxx>
Subject: Re:
Date: Sat, 10 Apr 1999 21:58:56 -0400
Jim Ducker wrote:
>
> RTs;
>
> I'm going to switch vendors for end-of-day historical and update
> data. I've used SuperCharts and played around a bit with TS. I did
a
> little back testing using the (limited) data provided with SC. I
used
> DialData for my EOD updating. I had trouble understanding Omega's
> recommendations for historical testing and optimizing, and some
time
> ago, I questioned Omega's people (including Bill Cruz) and Dial
Data.
> However, I didn’t get the information I was after before my poor
> health forced me to stop trading for a while. I'm starting up
again,
> and do NOT want to use DialData.
>
> I have paid attention to the general recommendations (and “stay
> aways”) on this and other forums and have a good idea of which
> vendor(s) I will consider. But first, I'm working on gaining a
better
> understanding (the advantages and disadvantages) of the various
types
> of historical data and roll over methodologies available from data
> providers. I want to check the group’s current consensus as to
which
> is the "best" method to use in SC/TS so as to most accurately
reflect
> how trading is actually done.
>
> My testing concerns include wanting to be able to roll over from
> (historical) contract to contract without trading each contract
> through its entire life (i.e. trade only the last months of each
near
> contract, then roll to the next at a reasonable point short of the
> FND or LTD). The data must, of course, be "adjusted", but again
which
> method is "best"? Proportional (Ratio Adjusted Data (RAD)? The
> pitches I've seen for that make a lot of sense to me. Do you folks
> agree?
>
I'm not an expert by any means but it seems for most purposes data
adjusted for the gap by means of addition/subtraction is better than
ratio, at least if ratio data changes the bar height.
Addn/subtraction
adjustment keeps the bar to bar and intrabar relationships the same
as
it was in real life. I am not that familiar with ratio adjusted but
it
seems that would change the height of the bar. This would change
your
P/L as well as the response of any part of the system based on
absolute
numbers (points or $ stops for example). Signals based on atr would
be
ok I believe. I think the one time addn/subtr does not work is if
you
system signals on %change, for example sell on a 2% drop.
Someone might argue that in a market that's been on an overall
uptrend
like the sp, you'd be better off with ratio adjusted bec that would
force the earlier bars to be more similar to todays. So you could
test
a system that did have fixed values (points or $) with older data.
But
what if the volatility of the sp changes in the future? I think
rather
than using ratio'd data, a system ought to be based on the atr, not
fixed points.
However, as I said i'm no systems expert, but rather a small trader
that
is not yet successful. So I would welcome corrections if needed.
> Omega warned that using data other than from their two "approved"
> sources would not be "seamless". As I understand it, this means
that
> I might have trouble “converting” others’ initial historical data
and
> then updating it with current EOD downloads so as to be able to
chart
> and test up-to-date daily, weekly or monthly bars. I gather it can
be
> done, but the way it is described seems unacceptably difficult to
do
> on a regular basis. I would like to hear from anyone with
experience
> with a “best” method who can tell me exactly how truly and fully
> "compatible" ("seamless") it is with SC/TS and from which vendor(s)
I
> can get it.
>
With vendors like CSI or Bridge, the data management is done outside
of
TS. You might have trouble trying to link a CSI historical contract
with Omega's current data, but if you did your daily downloading from
CSI you should have no trouble (at least with TS4).
I use Bridge. My impression is that Csi has wider variety in
continuous
contract maintanence and whether or not you include night data, but
has
had some glitches in their downloading software at times. Bridge has
been very reliable for downloading. I think it can do cont.
contracts
but I'm not sure.
Conrad Bowers
_______________________________________________________________
Get Free Email and Do More On The Web. Visit http://www.msn.com
|