|
PureBytes Links
Trading Reference Links
|
To follow up your posting, a question. I used to follow only futures, and
there it is usually clear
(1) when you have bad ticks (they're not frequent), and
(2) what you should do with them (throw them out)
I started following stocks (where I have no experience), and I see LOTS more
problem ticks:
(1) some are clearly small volume trades recorded late
(2) some are block trades (10-100x average size) sometime 1 point away
from other trades (and I wonder why the broker did not split the order)
There does not seem to be a one-size fits all solution. I am concerned
about:
(1) the effect of deleting ticks on stop and limit orders in RT trading
(2) the effect of deleting block trades on volume, which I would not
like altered.
So, I know technically how to clean them through synthetic bars, but
what seems the best thing to do? (if there is such a thing).
Thanks
----- Original Message -----
From: "Bilo Selhi" <biloselhi@xxxxxxxxxxx>
To: "Volker Knapp" <vk@xxxxxxxxxxxx>; "Beyond OddBall"
<beyond_oddball@xxxxxxxxx>; <omega-list@xxxxxxxxxx>
Sent: Saturday, March 16, 2002 6:04 PM
Subject: Re: A Professional comments on Pierre's software - - -
> you should know that even
> EOD data is mostly dirty... UNLESS you download
> it from a vendor that does TICK CLEANING and
> compiling of daily data every night after the market is
> closed.
> your claim that you can make money system trading
> on dirty data is, sorry to say, BS.
> most of pro systematic firms clean their data tick by
> tick and that's the only way to go.
> other considerations like saying you have to trade
> EOD because intraday data is dirty and because
> markets are manipulated is, sorry to say, BS.
> your claim to use the worst entry point is, sorry to say, BS.
>
> personal opinions don't count when trading, what counts
> is reality and the reality is in DATA:
> - you must clean data tick by tick to get clean intraday or
> end of day data. it is not hard to do, in fact it is pretty easy to
> accomplish if the platform is well designed...
> - your system must work on any time frame within certain
> data normality, liquidity, time constraints. EVEN 1 MIN
> AND TICK TIME FRAMES ALLOW PROFIT EXTRACTION
> IF... YOU KNOW HOW TO UTILIZE MARKET MICROSTRUCTURE.
> in fact there's been firms operating on the level of market microstructure
> for years now extracting millions of dollars on market scalping... and...
> there's been firms operating on monthly data and taking out millions
too...
> bottom line is, you can make it if you know how on any time frame...
> BUT NOT ON DIRTY DATA... systematic automated trading don't go
> hand in hand with bad data. it just does not fly...
>
> now, if we are talking primitive systems, or half ass systems or
methodologies
> or simply discretionary trading then who cares about how dirty the
> data is...
>
> CLEAN DATA + GOOD SYSTEM = PROFIT
> DIRTY DATA + GOOD SYSTEM = LOSS ( you can't take Ferrari off road , won't
work )
> CLEAN DATA + BAD SYSTEM = LOSS
> DIRTY DATA + BAD SYSTEM = LOSS
>
> note that the only way you make money is if you got both
> clean data and good system...
> THIS IS REALITY of systematic trading ...and it's simple.
>
> any trading platform that offers *system development and trading but
> does not address data issues = AMATEUR NIGHT.
> that's just about puts every existing *system development and
> trading platform into that category including tradestaion and your
> wealthlab platform, etc...
>
> bilo.
> ps. that's not just my opinion... it's the truth...
> also, good systems are never sold retail and clean rt data is hard to
find...
> you should understand why...look at the equation.
>
>
> ----- Original Message -----
> From: "Volker Knapp" <vk@xxxxxxxxxxxx>
> To: "Beyond OddBall" <beyond_oddball@xxxxxxxxx>; <omega-list@xxxxxxxxxx>
> Sent: Saturday, March 16, 2002 12:18 PM
> Subject: AW: A Professional comments on Pierre's software - - -
>
>
> > Hi,
> >
> > well in all fairness to PO and MB, this frustration results more from
> > someone who was not able to create a trading strategy with the given
> > software. That does not say anything about the quality of the
"developing"
> > software unless the software seller claims you will make huge profits
(which
> > sounds more like a system seller) with my software. Well, and maybe in
this
> > case it might be a border line.
> >
> > Anyway. I am doing system development since I can count numbers. Why was
I
> > doing it? To make easy money. Easy money for me is, place an order, be
> > percistent and make consistent profits. ;-)
> > Personally I feel that the focus of to many people here is on intra day
> > trading. Intra day trading is putting me infront of the screen again =
work
> > = no freedom = not for me!
> >
> > And then you get this never ending discussions of this RT data is 2
minutes
> > delayed, oh no it is not delayed, it is the time stamp that is delayed,
oh
> > no ticks are missing....!
> >
> > I say: So what!
> >
> > Any system that need 100% crystal clear intra day data will never
survive.
> > Because real live looks different. Then the quality of the real time
data
> > becomes the excuse for the personal failure (or failure of the system).
> > The market has been manipulated to do intra day trading (more
commission,
> > pay more for data, buy more expensive software). Or maybe those who
failed
> > in longterm trading think they have to make it short term? Who knows.
> >
> > Oh and did I mention commission - slippage - average profit ratio...
> >
> > There have been numerous articles in the www.traderclub.com forum, about
> > longterm trading with EOD data, good money management and diversified
> > markets. Some how it seems to work.
> > My suggestion is, test a system and use the worst entry point (the high
or
> > the low of the bar), and see how it would have done then.
> >
> > So, this is just my personal opinion.
> >
> > Volker Knapp
> > Wealth-Lab Inc.
> > http://www.wealth-lab.com
> > http://www.wealth-lab.de
> >
> >
> > PS: I am surprised to see PO and MB hand in hand ;-)
> >
> >
> >
> >
> >
> >
> > ++-----Ursprungliche Nachricht-----
> > ++Von: Beyond OddBall [mailto:beyond_oddball@xxxxxxxxx]
> > ++Gesendet: Samstag, 16. Marz 2002 15:12
> > ++An: omega-list@xxxxxxxxxx
> > ++Betreff: A Professional comments on Pierre's software - - -
> > ++
> > ++
> > ++
> > ++Before any of you go out and drop 29.5 G's (or even 5
> > ++G's) on any of Pierre's creations (or any other system
> > ++vendor) you may want to read one of the better emails
> > ++I received on Friday. This guy asked me to please not
> > ++mention specifically his former occupation (which
> > ++unfortunately would add even more validity to his
> > ++creditials). Note what he mentions though about his
> > ++previous career and you'll probably figure it out:
> > ++
> > ++
> > ++Beyond:
> > ++
> > ++I've been trying for years to find something that will
> > ++work Out Of Sample.
> > ++No luck; no big money lost so far, except Trade
> > ++Station costs.
> > ++
> > ++Spent 3 months one winter, trying to get Safir-X to
> > ++generate a trading system that would work OOS- no
> > ++cigar. In fairness to Pierre, he did
> > ++offer good technical support, and really tried to help
> > ++me succeed. However, I think Pierre's Back Prop NN
> > ++algorithm does not generalize well into
> > ++new, unseen market behavior (other NN algorithms may
> > ++work better, but this is not clearly supported by
> > ++research). But it was Au Revoir. Never saw anything
> > ++from Safir-X that came close to adequate OOS
> > ++performance.
> > ++
> > ++It was a demo, so I never got stuck paying the big
> > ++money Pierre is asking, for a product that did not
> > ++succeed for me.
> > ++
> > ++Spent the last 2 years trying to get various Trade
> > ++Station systems developed in several markets, but they
> > ++always failed in OOS tests.
> > ++
> > ++Worked with Neural Shell for 6 weeks: No luck. Got
> > ++most of my money back.
> > ++
> > ++Just spent 6 weeks with another trading software
> > ++package: No successful OOS performance that is even
> > ++remotely tradable. Supposed to get my money
> > ++back.
> > ++
> > ++And I've got programming skills, and scientific
> > ++research credentials/publications from a previous
> > ++career!
> > ++
> > ++I'm mad, and though I've been personally lucky so far,
> > ++I'm tired of hearing about the men and women who spent
> > ++big money on trading software, and ended up holding
> > ++the bag (a lot of my friends). No one should have to
> > ++spend money on trading software that is bogus.
> > ++
> > ++Now, after your post, I can get what I want cheap,
> > ++work with it for as long as I like, and if I ever do
> > ++get something that does work, I'll gladly
> > ++go to the vendor, and buy the latest version at the
> > ++retail price.
> > ++
> > ++Knowing that some sincere vendors do put in
> > ++considerable cost and effort into their products, my
> > ++conscience would push me to purchase the latest
> > ++version, at the current price. After I know the stuff
> > ++really works, and I can build a career on it.
> > ++
> > ++No more "30 day trials" that expire before I really
> > ++have had the time to thoroughly research and test out
> > ++trading software.
> > ++
> > ++Thanks from a grateful aspiring trader.
> > ++
> > ++********************************************
> > ++
> > ++Beyond sez:" Ok, Pierre, it's time to put up that
> > ++.ela, or shut up. Prove your stuff works. No more
> > ++beating around the bush, no more delays."
> > ++
> > ++
> >
> >
>
>
|