|
PureBytes Links
Trading Reference Links
|
> tj writes:
>
> >20-100 handles per what??? given any daily trading range, you can bang
> >this out in a single DAY trading only a coupla of times
>
> just to clarify, I was talking about a move in the SP of 20 to 100 "full"
> points.. i.e. if it goes from 1150.00 to 1250.00. This kind of trade lasts
> anywhere from a couple days to several weeks or more. Not sure if you'd
> call it a swing trade. Intermediate term? Something like that.
>
> Also interesting are the comments about STRESS. Whoever you are, it's a
> good bet that STRESS is directly proportional to the DOLLARS you are
> drawing down.
Stress? I've had my bypass surgery from dealing with my market prediction
technology --- but NOT from trading with it, but rather from the client
who initiated it and then wanted to own me ... :=}}
>
> Playing a longer time frame, the point moves for and against you will be
> LARGER, so all you have to do is trade proportionally SMALLER. If you do so
> the stress should be the same, no?
In my humble experience, the STRESS comes from the market moving in the
opposite direction to the system's predictions. And then BELIEVING that
the market will turn in the right direction and thus holding too long.
It takes a MECHANICAL belief in the system's predictions --- and I'm
not one who enjoys that type of trading. So, I'm starting to redevelop
my system to run every minute instead of every 15 minutes --- and then
to more carefully analyze the lead times of the indicators.
>
> All this aside, I'd love to trade faster. And take more advantage of the
> short-term volatility. The problem is, I have no good way to develop a good
> set of rules for daytrading. Can't find anything I can put into an EL
> system that's successful over a range of contracts. Maybe this or that
> procedure would work but unless I can backtest something to determine the
> odds, how can I trade it? Trail and error isn't the answer (I've tried).
>
For a year, I've run my predictive model every 15 minutes on 3 portfolios:
OEX, NDX, and DJIA. The predictive indicators offer (different) 60-minute,
90-minute, and 120 minute lead times. Some indicators are strong enough
to trade on their own; others require a "convergence" of indicators
at the same time in order to trade.
See http://apk.net/~lake for a summary of the system and its results.
The system also identifies individual stocks to play options on and an
indication of the risk associated with each trade. It also provides a
20- to 30-minute indication of when to exit the individual stock positions.
To reduce STRESS, I won't enter a position unless I can see the exit point
or adopt the strategy to exit at the first exit indication. My problem
is that I'm a technologist, not a trader, and I keep holding too long
and keep breaking the rules of day trading --- hoping for a better
position. Although I only do this when I have indicators that imply
the next-day's opening hours will move in my favor, I have gotten
killed by overnight events such as Brazil. Sure, the opening hours
move correctly, but the opening prices are so dramatically different
that the following movements are "right" but not good enough.
> TJ, you've been doing your thing for a long time so may I ask you a
> question?. Basically, how did you figure out your rules for daytrading in
> the first place? Did you piss away money trying out different theories, as
> I used to do? Maybe I just didn't hit the right thing soon enough!
>
Not even knowing what an option was, I started off (1) with some risk
management software tools I developed to measure health care quality
25 years ago and (2) a client who was running a hedge fund. Our concepts
behind modeling the markets is (I believe) based on a novel philosophy
--- but a technically/scientifically sound one. The intent was simply
to see if this one perspective would work, then to tune it if it did.
The fact of life was that the first week I turned it on,
it "saw" the major bear move of March 1996 and then got a life of its own.
It kept working on Overnight (closing
price) models and when we moved to intraday models I didn't have a clue
as to the predictive lead times of the various indicators. Took
a year to "see" what these times were. However, having a life of its
own, it took some serious real-money losses before I discovered an
operational bug for intraday models (Overnight models were OK). It
took another another 6 months of small gains and big losses to learn
(1) what day trading is about and (2) how strong the indicators had to
be to be highly reliable. It's not only having the indication of a move,
but what the risk vs. indicator strength is. Two months of following
the rules (see Web page above) convinced me:
A. The system isn't quite fast enough, but astoundingly accurate and
predictive;
B. The system requires me to cut and paste from Web portfolio quotes
into UNIX telnet windows and from UNIX telnet windows into
an Excel spreadsheet ALL DAY LONG. Getting to the bathroom
is difficult when trading seriously. This is not working because
I don't have time to think, especially when the Web data provider
feeds bum Open prices (blanks and zeroes), or when my ISP
cuts me off.
C. Trying to "push" the returns by taking greater risks is a losing
strategy. I don't have enough capital to make a living from
the system at 10% - 15% per month, and trying to increase the
return via risk has hurt. The system works perfectly, I'm just
NOT a trader and have psychological problems taking a loss to
limit losses.
Thus, I just stopped playing the system to do a major re-engineering of
the system to come close to running on every tick or every minute. I
always planned this, but the system
with a life-of-its-own became a monster-in-its-own-right!
The system, by the way, incorporates concepts from fuzzy logic to
quantify the risks of a given position; concepts from digital signal
processing to derive some 300 intermediate "indicators" from the simple
Open High Low Last Volume data that are the only inputs; and is
written in the C languages and runs on essentially any hardware (although
Mac and UNIX based now).
One suggestion to those of you who are building or buying or licensing
systems. From my long background in digital signal processing:
First run a "step function" and an "impulse function" through the system.
Just create a fake data file with, say, 50 days of fake history with
nothing changing but the Open price at day 25 jumping from $10 to $100.
See what the various calculated indicators from the system do. In one
system (not ours), the indicators went ballistic or unstable 72 days
later! The "impulse function " would have the Open jump from $10
to $100 and back to $10. The initial client had claiming this candidate
system was "yanking me around". No wonder! Because of this, I decided
to "roll my own".
I'll be back in a few months after the re-engineering is complete, with
more specific details.
This discussions in this group have been great for the few weeks since
I joined! Thank you all!
> sincerely, phil
Cheers,
Rob Lake
rbl@xxxxxxxxxxx
|