|
PureBytes Links
Trading Reference Links
|
John --
Appreciate the info on Neoticker, but you only confirmed it
is not yet ready for prime-time. I'm interested in
heavy-duty back-testing, which, includes optimization.
I have a Visual C++ software program I'm developing with a
friend that, thought not finished, is able to optimize, say
500 iterations on a few hundred stocks, at the rate of 1/4
second per 10-year's daily data. (Intra-day data is just a
multiple.) Results are saved in a Grid/database that
permits sorting on any variable, or output stat I choose.
I've also just ported my old Dos Excalibur program
(customized version of Futures Truth's testing platform), to
Linux, which runs about 17 times faster than above, or 0.014
second per 10-year's daily data. I had used the Dos ver for
futures testing for 7-8 years, but it can't handle massive
stock data.
Both programs are already geared to handle 100 variables;
create basic portfolios (still developing, and initially
I'll probably use a new Rina product that can handle output
from non-Trade Station products.) FYI: I don't use 100
variables, but rather have say, 10-15 different variables,
for 3-5 different exits. It's just easier to keep a master
grid with all the variables, rather than juggle them for
each system. Then I just toggle on/off which
entries/exits/variables I want for a particular test.
I suspect your idea of taking each system's output from
NeoTicker into Excel might make sense for a few futures
contracts, but would exceed my current 1/4 second per
10-year's data throughput when testing hundreds of
iterations on hundreds of stocks.
I'm now working on porting my C++ charting program, which
integrates with Excalibur, to Linux as well. There I can
select any of the hundreds of results of output, for any
stock, & chart on daily graph (even though the test was on
intra-day data) and see the standard buy/sell points, equity
curve,etc.; as well as a cursor that goes across 4 panes
with plotted charts (useful when analyzing the same signal
with different variables.)
Fortunately, I have full source code for both of the above
programs, so, with the help of programmers, I can get done
what I want; deal with bugs, etc. The Visual C++ platform
has already taken longer than I expected (1 year, not 12+),
but the delay was mostly due to my being involved in trading
rather than paying attention to what my programmer was
doing.
I'm working on the above programs for huge stock testing.
Perhaps Neoticker could handle a more limited universe.
I mention the above for info purposes only. I have no plans
or desires toward marketing the software. I know how to
trade; can direct programmers to get done what I need; but
selling is neither an interest nor ability I wish to
cultivate. Besides, as the omega-list demonstrates, to
create a product to satisfy even a relatively small number
of users is a huge task. I only need to accomplish my own
goals. I should add, the expense is rather considerable,
about $ 150,000 so far on the Visual C++ platform; the
Linux port so far has nominal costs, but I'd guess another $
10 - $30,000 over the next few months.
However, I'm interested in seeing what is available, & if I
found a product that could do what I want now, I would buy
it.
Thanks again for the info. I'd be interested to hear, as
things change at NeoTicker.
Regards,
----- Original Message -----
From: "John O. Romero" <jromero@xxxxxxxxx>
To: "Michael Berger" <mberger@xxxxxxxx>; ".OmegaList"
<omega-list@xxxxxxxxxx>
Sent: Wednesday, August 29, 2001 10:05 AM
Subject: Re: NeoTicker (was Re: EL to VB translator)
> Michael,
>
> I have placed my replies your question after each. Hope it
helps. By the
> way, you can see the entire documentation online if you
want to.
>
> At 8/29/2001 07:57 AM, Michael Berger wrote:
> >Following-up from a message I posted here in July:
> >***********************************
> >Just spent a few minutes at www.neoticker.com
> >
> >1. They seem to have limited backtesting, & no
> >optimization.
> They have full backtesting capabilities, but do not have
any optimization
> yet. You would have to code this yourself. Not tirvial,
but not rocket
> science. When writing your systems, there is not limit to
the combinations
> of what you test. you can have the system based on one
data steam and
> buy/sell another, or an entire basket of symbols.
>
>
> >2. There are a max of 8 variables.
>
> The limit of 8 input variable will be removed in the next
point
> release. Supposed to be out in mid-late Sept. At that
point it will be
> unlimited.
>
> >3. They use a scripting language, as wealth-lab does, so
it
> >might also have speed issues.
> It does use scripting languages. That was the first set
of indicator
> support they began with. Since then, they now support
compiled code -
> either ActiveX Exe or .DLL. They have examples of this
and also show how
> to use someone elses code. The example is for Jurik
indicators.
>
>
> >**************************
> >Another member said it does not have portfolio
capabilities.
>
>
> There is no portfolio capabilities at this time. I use a
seperate program
> that also uses the M3 as the data server to keep realtime
updates. If all
> you want to do is track your current investments, there
are alert windows
> as well as report windows. Also, there is a crude type of
portfolio
> tracking capability. You can use the Quote window to list
multiple symbols
> you want to track, high, low, etc. One of the fields is
position. You can
> input a quantity, cost and multiplier. Then you also have
the ability to
> sum your positions. So there is a minimal amount of
portifolio
> tracking. Not a full fledged portfolio program.
>
> John
|