Hello,

Does anybody know how I gave to implement a trailingstop which starts
at a certain profit level? Say when de price is 10% above my
buyingprice the trailingstop needs to start working.

I really don't know how to fix this, so I will be very thankful for
every command that will help me further.

Greets Ronald

Thanks for the reply guys!

Are there any samples I can look at before I start working on my own
custom backtesting script?

thanks for the help


--- In amibroker-ts@yahoogroups.com, "Terry" <MagicTH@...> wrote:
>
> You may also want to use temporary Buy-Sell-Short-Cover variables
such
> as myBuy or newBuy inside the loop. When the loop is finished set:
>
> Buy = newBuy;
>
> Sell = newSell;
>
> Etc.
>
> --
>
> Terry
>
> -----Original Message-----
> From: amibroker-ts@yahoogroups.com [mailto:amibroker-
ts@yahoogroups.com]
> On Behalf Of Graham
> Sent: Sunday, August 27, 2006 16:38
> To: amibroker-ts@yahoogroups.com
> Subject: Re: [amibroker-ts] Problems when zooming in
>
>
>
> when using looping you may need to add this line to your code
> setbarsrequired(10000,10000);
> the number of bars toi suit your data length
> --
> Cheers
> Graham
> AB-Write >< Professional AFL Writing Service
> Yes, I write AFL code to your requirements
> http://e-wire. <http://e-wire.net.au/~eb_kavan/ab_write.htm>
> net.au/~eb_kavan/ab_write.htm
>
> On 27/08/06, da_pbear <da_pbear@yahoo. <mailto:da_pbear%
40yahoo.com>
> com> wrote:
> > I have a trading system that uses a loop methodology in its
> > implimentation.
> >
> > Ie.
> >
> > <..>
> >
> > For(i = 100; i < BarCount; i++)
> > {
> > //Stuff
> > }
> >
> > </..>
> >
> > It generates all the buy and sell signals within this for loop. I
> > have been a Wealth lab user and are now only wondering into the
> > world of Amibroker.
> >
> > One of the issue which I have been encountering is the "re
> > evaluation" of my signals when I zoom in/out of the chart
setting.
> > Looks like the BarCount changes everytime I zoom in or zoom out.
I
> > do not want that. My signals are regenerated and different
everytime
> > I zoom in or out. I want the chart to be calculated based on the
> > number of Bars defined in the defaults and not "redraw" itself
> > everytime I zoom in and out. Is there a way to correct this?
Could
> > this be due to the fact that I am using a loop based approach
> > compared to the regular BarbyBar calculation based approach.
> >
> > 2nd issue: My system is martingale complaint. AmiBroker
backtester
> > doesnt seem to recongnize this fact. Is there a way to implement
> > this in the backtesting process apart from writing my own
> > backtesting script?
> >
> > Thanks and appreciate the help!
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> > Yahoo! Groups Links
> >
> >
> >
> >
> >
> >
> >
> >
>
>
>
>
>
> [Non-text portions of this message have been removed]
>

Hello,
I have the same problem. The applystop formula stops a position after
x bars but not after x days AND a profit minor then x percent.  How do
I code this?


--- In amibroker-ts@yahoogroups.com, "Terry" <MagicTH@...> wrote:
>
> ApplyStop and using nBarStop.
>
>
>
> --
>
> Terry
>
> -----Original Message-----
> From: amibroker-ts@yahoogroups.com [mailto:amibroker-ts@yahoogroups.com]
> On Behalf Of Kal Parekh
> Sent: Thursday, July 20, 2006 13:40
> To: amibroker-ts@yahoogroups.com
> Subject: [amibroker-ts] Timed Exit
>
>
>
> Hi All,
>
> Thank you for time and support.
>
> Quick ques: Can anyone point to AB code that allows me simulate exit
> after n-bars if a position is unprofitable? i.e, exit trade after 10
> days is the trade is unprofitable.
>
> Again, thank you for your time and support.
>
> KP
>
>
>
>
>
> [Non-text portions of this message have been removed]
>

See HELP for designating % or nBar. I have not tried using TWO ApplyStop
statements simultaneously. I would not be surprised if that would work.
Use one for % and the 2nd for days using nBar * no.bars per day
depending on your time frame. The last part might get a little tricky if
you enter at, say 3PM, but want to get out on the morning of the 2nd
day. You would have to compute the number of bars until that time.and
check the option for variable stops.



If the above doesn't work, then you will have to write your own stops,
probably in a loop, and you will have to track your status and P&L. You
actually get more control this way and it's clearer (to me) as to what
is really happening. (Note that different types of stops get assigned
different values of Sell/Cover. 1 = a normal exit. 2 = a stop loss. 3 =
a profit stop. Etc.



Possibly this could also be done in the Custom Backtester, but that
involves another level of learning.

--

Terry

-----Original Message-----
From: amibroker-ts@yahoogroups.com [mailto:amibroker-ts@yahoogroups.com]
On Behalf Of supistarde
Sent: Saturday, September 16, 2006 00:30
To: amibroker-ts@yahoogroups.com
Subject: [amibroker-ts] Re: Timed Exit



Hello,
I have the same problem. The applystop formula stops a position after
x bars but not after x days AND a profit minor then x percent. How do
I code this?

--- In amibroker-ts@ <mailto:amibroker-ts%40yahoogroups.com>
yahoogroups.com, "Terry" <MagicTH@...> wrote:
>
> ApplyStop and using nBarStop.
>
>
>
> --
>
> Terry
>
> -----Original Message-----
> From: amibroker-ts@ <mailto:amibroker-ts%40yahoogroups.com>
yahoogroups.com [mailto:amibroker-ts@
<mailto:amibroker-ts%40yahoogroups.com> yahoogroups.com]
> On Behalf Of Kal Parekh
> Sent: Thursday, July 20, 2006 13:40
> To: amibroker-ts@ <mailto:amibroker-ts%40yahoogroups.com>
yahoogroups.com
> Subject: [amibroker-ts] Timed Exit
>
>
>
> Hi All,
>
> Thank you for time and support.
>
> Quick ques: Can anyone point to AB code that allows me simulate exit
> after n-bars if a position is unprofitable? i.e, exit trade after 10
> days is the trade is unprofitable.
>
> Again, thank you for your time and support.
>
> KP
>
>
>
>
>
> [Non-text portions of this message have been removed]
>





[Non-text portions of this message have been removed]

I am a big fan of Monte Carlo Simulation (MCS), so as much as
backtesting makes sense, I don't think it really means much unless
you followup your results with MCS to make sure your results are
real.

One idea is to optimize your system against the total of all symbols
that you are interested in trading.

Then, trade those settings against each individual symbol, one at a
time. Take each results set for each symbol, and check it against
MCS. The best performers will be the ones with the best MCS ratios.

I am not a statistician, although my experience with backtesting and
MCS is that MCS is profoundly more reliable in its results. MCS was
originally developed to predict where nuclear fallout would end up,
on and around nuclear test ranges. It was highly accurate from the
start of its existence.

I would LOVE to see automatic MCS integration with AB. If anyone has
a script for doing this within AB, or in conjunction with an Excel
doc that handles the MCS, that would be fantastic!! I already have
an Excel doc for handling the MCS. Just would like to automate the
ongoing system backtesting against MCS, that's all.

Thanks,

Brian

--- In amibroker-ts@yahoogroups.com, Sidney Kaiser <s9kaiser@...>
wrote:
>
> I was doing a little "simple", or so I thought, trading system
work this
> weekend and was reminded once again how it is not simple.
>
> System: StochRSI fixed trigger levels, appx buy 20 sell 80
> 	 uses a trend filter appx 25-45 SMA, only take trades
in the direction of
> the trend
> 	 trailing stop
> 	 long only, daily data
> 	 start date 1/2/2003
>
> Stock list:  I started with the One Chicago list of single stock
futures,
> appx 120 issues.
> 	 ran a coarse optimization of params to find which
stocks would trade well
> with StochRSI
> 	 created a new watch list with the issues that had >
10% profit over the
> test period
> 	 using the old V4.4 backtester to quickly generate
stock by stock results,
> 58 stocks
>
> I followed that up with a final optimization of params on the 58
stock list
> then ran the default portfolio backtester.  Starting with $10k the
profit
> was $68.6k.
>
> I then changed the stock list slightly and ended up with 62
stocks, the
> portfolio BT profit was $32.7k.
>
> Changing the stock list selection to > 20% profit and re-testing
with 49
> stocks, the portfolio BT profit was $24.4k.
>
> At first I thought, this was weird, but then I remembered
something I first
> learned years ago when doing backtests with FastTools.  The
portfolio
> tester is choosing one stock at a time from the list of available
> stocks.  If I slightly change the composition of the list, a new,
unique,
> sequence of trades is generated with very different results.  This
is
> something that I need to keep in the forefront of my thinking when
I am
> evaluating backtest results.
>
> It is very difficult to evaluate the effects of changing the list
of stocks
> because of this sequencing of trades effect.
>
> At this point I am faced with doing a full blown position score
approach to
> the backtesting.  That is unfortunate because I have never been
very
> successful with devising useful position scoring routines.
>
> Am I overlooking a simpler approach?
>
> If I have to go this way, what sort of position score  code might
work out
> here?
>
> TIA
> Sid
>
>
> ---
> Outgoing mail is certified Virus Free.
> Checked by AVG anti-virus system (http://www.grisoft.com).
> Version: 6.0.762 / Virus Database: 510 - Release Date: 9/13/2004
>

So its success with nuclear fallout gives it credence?

I'd limit Monte Carlo calcs to mathematical problems that I couldn't [or was
too lazy to] analyze.

Bob

-----Original Message-----
From: amibroker-ts@yahoogroups.com
[mailto:amibroker-ts@yahoogroups.com]On Behalf Of brpnw1
Sent: Monday, September 18, 2006 6:31 PM
To: amibroker-ts@yahoogroups.com
Subject: [amibroker-ts] Re: trading system subtleties, level 1


I am a big fan of Monte Carlo Simulation (MCS), so as much as
backtesting makes sense, I don't think it really means much unless
you followup your results with MCS to make sure your results are
real.

One idea is to optimize your system against the total of all symbols
that you are interested in trading.

Then, trade those settings against each individual symbol, one at a
time. Take each results set for each symbol, and check it against
MCS. The best performers will be the ones with the best MCS ratios.

I am not a statistician, although my experience with backtesting and
MCS is that MCS is profoundly more reliable in its results. MCS was
originally developed to predict where nuclear fallout would end up,
on and around nuclear test ranges. It was highly accurate from the
start of its existence.

I would LOVE to see automatic MCS integration with AB. If anyone has
a script for doing this within AB, or in conjunction with an Excel
doc that handles the MCS, that would be fantastic!! I already have
an Excel doc for handling the MCS. Just would like to automate the
ongoing system backtesting against MCS, that's all.

Thanks,

Brian

--- In amibroker-ts@yahoogroups.com, Sidney Kaiser <s9kaiser@...>
wrote:
>
> I was doing a little "simple", or so I thought, trading system
work this
> weekend and was reminded once again how it is not simple.
>
> System: StochRSI fixed trigger levels, appx buy 20 sell 80
> 	 uses a trend filter appx 25-45 SMA, only take trades
in the direction of
> the trend
> 	 trailing stop
> 	 long only, daily data
> 	 start date 1/2/2003
>
> Stock list:  I started with the One Chicago list of single stock
futures,
> appx 120 issues.
> 	 ran a coarse optimization of params to find which
stocks would trade well
> with StochRSI
> 	 created a new watch list with the issues that had >
10% profit over the
> test period
> 	 using the old V4.4 backtester to quickly generate
stock by stock results,
> 58 stocks
>
> I followed that up with a final optimization of params on the 58
stock list
> then ran the default portfolio backtester.  Starting with $10k the
profit
> was $68.6k.
>
> I then changed the stock list slightly and ended up with 62
stocks, the
> portfolio BT profit was $32.7k.
>
> Changing the stock list selection to > 20% profit and re-testing
with 49
> stocks, the portfolio BT profit was $24.4k.
>
> At first I thought, this was weird, but then I remembered
something I first
> learned years ago when doing backtests with FastTools.  The
portfolio
> tester is choosing one stock at a time from the list of available
> stocks.  If I slightly change the composition of the list, a new,
unique,
> sequence of trades is generated with very different results.  This
is
> something that I need to keep in the forefront of my thinking when
I am
> evaluating backtest results.
>
> It is very difficult to evaluate the effects of changing the list
of stocks
> because of this sequencing of trades effect.
>
> At this point I am faced with doing a full blown position score
approach to
> the backtesting.  That is unfortunate because I have never been
very
> successful with devising useful position scoring routines.
>
> Am I overlooking a simpler approach?
>
> If I have to go this way, what sort of position score  code might
work out
> here?
>
> TIA
> Sid
>
>
> ---
> Outgoing mail is certified Virus Free.
> Checked by AVG anti-virus system (http://www.grisoft.com).
> Version: 6.0.762 / Virus Database: 510 - Release Date: 9/13/2004
>










Yahoo! Groups Links

Personally I'm more interested in seeing out of sample
performance ...

--- In amibroker-ts@yahoogroups.com, "brpnw1" <tradermail@...> wrote:
>
> I am a big fan of Monte Carlo Simulation (MCS), so as much as
> backtesting makes sense, I don't think it really means much unless
> you followup your results with MCS to make sure your results are
> real.
>
> One idea is to optimize your system against the total of all
symbols
> that you are interested in trading.
>
> Then, trade those settings against each individual symbol, one at a
> time. Take each results set for each symbol, and check it against
> MCS. The best performers will be the ones with the best MCS ratios.
>
> I am not a statistician, although my experience with backtesting
and
> MCS is that MCS is profoundly more reliable in its results. MCS was
> originally developed to predict where nuclear fallout would end up,
> on and around nuclear test ranges. It was highly accurate from the
> start of its existence.
>
> I would LOVE to see automatic MCS integration with AB. If anyone
has
> a script for doing this within AB, or in conjunction with an Excel
> doc that handles the MCS, that would be fantastic!! I already have
> an Excel doc for handling the MCS. Just would like to automate the
> ongoing system backtesting against MCS, that's all.
>
> Thanks,
>
> Brian
>
> --- In amibroker-ts@yahoogroups.com, Sidney Kaiser <s9kaiser@>
> wrote:
> >
> > I was doing a little "simple", or so I thought, trading system
> work this
> > weekend and was reminded once again how it is not simple.
> >
> > System: StochRSI fixed trigger levels, appx buy 20 sell 80
> > 	 uses a trend filter appx 25-45 SMA, only take trades
> in the direction of
> > the trend
> > 	 trailing stop
> > 	 long only, daily data
> > 	 start date 1/2/2003
> >
> > Stock list:  I started with the One Chicago list of single stock
> futures,
> > appx 120 issues.
> > 	 ran a coarse optimization of params to find which
> stocks would trade well
> > with StochRSI
> > 	 created a new watch list with the issues that had >
> 10% profit over the
> > test period
> > 	 using the old V4.4 backtester to quickly generate
> stock by stock results,
> > 58 stocks
> >
> > I followed that up with a final optimization of params on the 58
> stock list
> > then ran the default portfolio backtester.  Starting with $10k
the
> profit
> > was $68.6k.
> >
> > I then changed the stock list slightly and ended up with 62
> stocks, the
> > portfolio BT profit was $32.7k.
> >
> > Changing the stock list selection to > 20% profit and re-testing
> with 49
> > stocks, the portfolio BT profit was $24.4k.
> >
> > At first I thought, this was weird, but then I remembered
> something I first
> > learned years ago when doing backtests with FastTools.  The
> portfolio
> > tester is choosing one stock at a time from the list of available
> > stocks.  If I slightly change the composition of the list, a new,
> unique,
> > sequence of trades is generated with very different results.
This
> is
> > something that I need to keep in the forefront of my thinking
when
> I am
> > evaluating backtest results.
> >
> > It is very difficult to evaluate the effects of changing the list
> of stocks
> > because of this sequencing of trades effect.
> >
> > At this point I am faced with doing a full blown position score
> approach to
> > the backtesting.  That is unfortunate because I have never been
> very
> > successful with devising useful position scoring routines.
> >
> > Am I overlooking a simpler approach?
> >
> > If I have to go this way, what sort of position score  code might
> work out
> > here?
> >
> > TIA
> > Sid
> >
> >
> > ---
> > Outgoing mail is certified Virus Free.
> > Checked by AVG anti-virus system (http://www.grisoft.com).
> > Version: 6.0.762 / Virus Database: 510 - Release Date: 9/13/2004
> >
>

Brian --
I have a problem with the following statement, "optimize your system
against the total of all symbols that you are interested in trading.
Then, trade those settings against each individual symbol, one at a time."

Say your interested in trading stocks in the NASDAQ and therefore
optimize on the NASDAQ.  A majority of the stocks in the NASDAQ are
highly correlated with it and therefore your system has a significant
bias toward those stocks and to the _*time frame*_ of your
optimization.  A rising tide, etc.

I don't really know much about Monte Carlo, except as it might be
applied to something relatively simple like calculating probable system
performance, failure rates, etc. in electrical engineering.  Perhaps you
would enlighten us on how it could be applied to our trading systems?
"The devil is in the Details."
-- Keith

brpnw1 wrote:
>
> I am a big fan of Monte Carlo Simulation (MCS), so as much as
> backtesting makes sense, I don't think it really means much unless
> you followup your results with MCS to make sure your results are
> real.
>
> One idea is to optimize your system against the total of all symbols
> that you are interested in trading.
>
> Then, trade those settings against each individual symbol, one at a
> time. Take each results set for each symbol, and check it against
> MCS. The best performers will be the ones with the best MCS ratios.
>
> I am not a statistician, although my experience with backtesting and
> MCS is that MCS is profoundly more reliable in its results. MCS was
> originally developed to predict where nuclear fallout would end up,
> on and around nuclear test ranges. It was highly accurate from the
> start of its existence.
>
> I would LOVE to see automatic MCS integration with AB. If anyone has
> a script for doing this within AB, or in conjunction with an Excel
> doc that handles the MCS, that would be fantastic!! I already have
> an Excel doc for handling the MCS. Just would like to automate the
> ongoing system backtesting against MCS, that's all.
>
> Thanks,
>
> Brian
>
> --- In amibroker-ts@yahoogroups.com
> <mailto:amibroker-ts%40yahoogroups.com>, Sidney Kaiser <s9kaiser@...>
> wrote:
> >
> > I was doing a little "simple", or so I thought, trading system
> work this
> > weekend and was reminded once again how it is not simple.
> >
> > System: StochRSI fixed trigger levels, appx buy 20 sell 80
> > uses a trend filter appx 25-45 SMA, only take trades
> in the direction of
> > the trend
> > trailing stop
> > long only, daily data
> > start date 1/2/2003
> >
> > Stock list: I started with the One Chicago list of single stock
> futures,
> > appx 120 issues.
> > ran a coarse optimization of params to find which
> stocks would trade well
> > with StochRSI
> > created a new watch list with the issues that had >
> 10% profit over the
> > test period
> > using the old V4.4 backtester to quickly generate
> stock by stock results,
> > 58 stocks
> >
> > I followed that up with a final optimization of params on the 58
> stock list
> > then ran the default portfolio backtester. Starting with $10k the
> profit
> > was $68.6k.
> >
> > I then changed the stock list slightly and ended up with 62
> stocks, the
> > portfolio BT profit was $32.7k.
> >
> > Changing the stock list selection to > 20% profit and re-testing
> with 49
> > stocks, the portfolio BT profit was $24.4k.
> >
> > At first I thought, this was weird, but then I remembered
> something I first
> > learned years ago when doing backtests with FastTools. The
> portfolio
> > tester is choosing one stock at a time from the list of available
> > stocks. If I slightly change the composition of the list, a new,
> unique,
> > sequence of trades is generated with very different results. This
> is
> > something that I need to keep in the forefront of my thinking when
> I am
> > evaluating backtest results.
> >
> > It is very difficult to evaluate the effects of changing the list
> of stocks
> > because of this sequencing of trades effect.
> >
> > At this point I am faced with doing a full blown position score
> approach to
> > the backtesting. That is unfortunate because I have never been
> very
> > successful with devising useful position scoring routines.
> >
> > Am I overlooking a simpler approach?
> >
> > If I have to go this way, what sort of position score code might
> work out
> > here?
> >
> > TIA
> > Sid
> >
> >
> > ---
> > Outgoing mail is certified Virus Free.
> > Checked by AVG anti-virus system (http://www.grisoft.com
> <http://www.grisoft.com>).
> > Version: 6.0.762 / Virus Database: 510 - Release Date: 9/13/2004
> >
>
>


[Non-text portions of this message have been removed]

You say that as if MCS is somehow a questionable or fringe method.  It
is not.  It is used in some capacity by probably every large financial
institution in the world, and has tons of applications in other fields
as well, some of which are listed here.

http://en.wikipedia.org/wiki/Monte_Carlo_method

--- In amibroker-ts@yahoogroups.com, "Bob Jagow" <bjagow@...> wrote:
>
> So its success with nuclear fallout gives it credence?
>
> I'd limit Monte Carlo calcs to mathematical problems that I couldn't
[or was
> too lazy to] analyze.
>
> Bob
>
> -----Original Message-----
> From: amibroker-ts@yahoogroups.com
> [mailto:amibroker-ts@yahoogroups.com]On Behalf Of brpnw1
> Sent: Monday, September 18, 2006 6:31 PM
> To: amibroker-ts@yahoogroups.com
> Subject: [amibroker-ts] Re: trading system subtleties, level 1
>
>
> I am a big fan of Monte Carlo Simulation (MCS), so as much as
> backtesting makes sense, I don't think it really means much unless
> you followup your results with MCS to make sure your results are
> real.
>
> One idea is to optimize your system against the total of all symbols
> that you are interested in trading.
>
> Then, trade those settings against each individual symbol, one at a
> time. Take each results set for each symbol, and check it against
> MCS. The best performers will be the ones with the best MCS ratios.
>
> I am not a statistician, although my experience with backtesting and
> MCS is that MCS is profoundly more reliable in its results. MCS was
> originally developed to predict where nuclear fallout would end up,
> on and around nuclear test ranges. It was highly accurate from the
> start of its existence.
>
> I would LOVE to see automatic MCS integration with AB. If anyone has
> a script for doing this within AB, or in conjunction with an Excel
> doc that handles the MCS, that would be fantastic!! I already have
> an Excel doc for handling the MCS. Just would like to automate the
> ongoing system backtesting against MCS, that's all.
>
> Thanks,
>
> Brian
>
> --- In amibroker-ts@yahoogroups.com, Sidney Kaiser <s9kaiser@>
> wrote:
> >
> > I was doing a little "simple", or so I thought, trading system
> work this
> > weekend and was reminded once again how it is not simple.
> >
> > System: StochRSI fixed trigger levels, appx buy 20 sell 80
> > 	 uses a trend filter appx 25-45 SMA, only take trades
> in the direction of
> > the trend
> > 	 trailing stop
> > 	 long only, daily data
> > 	 start date 1/2/2003
> >
> > Stock list:  I started with the One Chicago list of single stock
> futures,
> > appx 120 issues.
> > 	 ran a coarse optimization of params to find which
> stocks would trade well
> > with StochRSI
> > 	 created a new watch list with the issues that had >
> 10% profit over the
> > test period
> > 	 using the old V4.4 backtester to quickly generate
> stock by stock results,
> > 58 stocks
> >
> > I followed that up with a final optimization of params on the 58
> stock list
> > then ran the default portfolio backtester.  Starting with $10k the
> profit
> > was $68.6k.
> >
> > I then changed the stock list slightly and ended up with 62
> stocks, the
> > portfolio BT profit was $32.7k.
> >
> > Changing the stock list selection to > 20% profit and re-testing
> with 49
> > stocks, the portfolio BT profit was $24.4k.
> >
> > At first I thought, this was weird, but then I remembered
> something I first
> > learned years ago when doing backtests with FastTools.  The
> portfolio
> > tester is choosing one stock at a time from the list of available
> > stocks.  If I slightly change the composition of the list, a new,
> unique,
> > sequence of trades is generated with very different results.  This
> is
> > something that I need to keep in the forefront of my thinking when
> I am
> > evaluating backtest results.
> >
> > It is very difficult to evaluate the effects of changing the list
> of stocks
> > because of this sequencing of trades effect.
> >
> > At this point I am faced with doing a full blown position score
> approach to
> > the backtesting.  That is unfortunate because I have never been
> very
> > successful with devising useful position scoring routines.
> >
> > Am I overlooking a simpler approach?
> >
> > If I have to go this way, what sort of position score  code might
> work out
> > here?
> >
> > TIA
> > Sid
> >
> >
> > ---
> > Outgoing mail is certified Virus Free.
> > Checked by AVG anti-virus system (http://www.grisoft.com).
> > Version: 6.0.762 / Virus Database: 510 - Release Date: 9/13/2004
> >
>
>
>
>
>
>
>
>
>
>
> Yahoo! Groups Links
>

Does anyone have any ideas on how difficult it is to predict the
position of the Open and Close relative to each other.  For example an
Open at 650 and a Close of 655
would be an up day, ignoring pervious market activity. The goal here
is a simple day trading system that I don't have to worry about while
at work, just buy or sell at Open and exit at Close.

Any ideas on the accuracy ratio of published systems or advisory
services in doing this?

GFN

It's really a Greek thing, 714, but I think further ranting should await
responses from a modicum of the other 713 quanttraders :)

-----Original Message-----
From: amibroker-ts@yahoogroups.com
[mailto:amibroker-ts@yahoogroups.com]On Behalf Of quanttrader714
Sent: Wednesday, September 20, 2006 11:49 AM
To: amibroker-ts@yahoogroups.com
Subject: [amibroker-ts] Re: trading system subtleties, level 1


You say that as if MCS is somehow a questionable or fringe method.  It
is not.  It is used in some capacity by probably every large financial
institution in the world, and has tons of applications in other fields
as well, some of which are listed here.

http://en.wikipedia.org/wiki/Monte_Carlo_method

--- In amibroker-ts@yahoogroups.com, "Bob Jagow" <bjagow@...> wrote:
>
> So its success with nuclear fallout gives it credence?
>
> I'd limit Monte Carlo calcs to mathematical problems that I couldn't
[or was
> too lazy to] analyze.
>
> Bob
>
> -----Original Message-----
> From: amibroker-ts@yahoogroups.com
> [mailto:amibroker-ts@yahoogroups.com]On Behalf Of brpnw1
> Sent: Monday, September 18, 2006 6:31 PM
> To: amibroker-ts@yahoogroups.com
> Subject: [amibroker-ts] Re: trading system subtleties, level 1
>
>
> I am a big fan of Monte Carlo Simulation (MCS), so as much as
> backtesting makes sense, I don't think it really means much unless
> you followup your results with MCS to make sure your results are
> real.
>
> One idea is to optimize your system against the total of all symbols
> that you are interested in trading.
>
> Then, trade those settings against each individual symbol, one at a
> time. Take each results set for each symbol, and check it against
> MCS. The best performers will be the ones with the best MCS ratios.
>
> I am not a statistician, although my experience with backtesting and
> MCS is that MCS is profoundly more reliable in its results. MCS was
> originally developed to predict where nuclear fallout would end up,
> on and around nuclear test ranges. It was highly accurate from the
> start of its existence.
>
> I would LOVE to see automatic MCS integration with AB. If anyone has
> a script for doing this within AB, or in conjunction with an Excel
> doc that handles the MCS, that would be fantastic!! I already have
> an Excel doc for handling the MCS. Just would like to automate the
> ongoing system backtesting against MCS, that's all.
>
> Thanks,
>
> Brian
>
> --- In amibroker-ts@yahoogroups.com, Sidney Kaiser <s9kaiser@>
> wrote:
> >
> > I was doing a little "simple", or so I thought, trading system
> work this
> > weekend and was reminded once again how it is not simple.
> >
> > System: StochRSI fixed trigger levels, appx buy 20 sell 80
> > 	 uses a trend filter appx 25-45 SMA, only take trades
> in the direction of
> > the trend
> > 	 trailing stop
> > 	 long only, daily data
> > 	 start date 1/2/2003
> >
> > Stock list:  I started with the One Chicago list of single stock
> futures,
> > appx 120 issues.
> > 	 ran a coarse optimization of params to find which
> stocks would trade well
> > with StochRSI
> > 	 created a new watch list with the issues that had >
> 10% profit over the
> > test period
> > 	 using the old V4.4 backtester to quickly generate
> stock by stock results,
> > 58 stocks
> >
> > I followed that up with a final optimization of params on the 58
> stock list
> > then ran the default portfolio backtester.  Starting with $10k the
> profit
> > was $68.6k.
> >
> > I then changed the stock list slightly and ended up with 62
> stocks, the
> > portfolio BT profit was $32.7k.
> >
> > Changing the stock list selection to > 20% profit and re-testing
> with 49
> > stocks, the portfolio BT profit was $24.4k.
> >
> > At first I thought, this was weird, but then I remembered
> something I first
> > learned years ago when doing backtests with FastTools.  The
> portfolio
> > tester is choosing one stock at a time from the list of available
> > stocks.  If I slightly change the composition of the list, a new,
> unique,
> > sequence of trades is generated with very different results.  This
> is
> > something that I need to keep in the forefront of my thinking when
> I am
> > evaluating backtest results.
> >
> > It is very difficult to evaluate the effects of changing the list
> of stocks
> > because of this sequencing of trades effect.
> >
> > At this point I am faced with doing a full blown position score
> approach to
> > the backtesting.  That is unfortunate because I have never been
> very
> > successful with devising useful position scoring routines.
> >
> > Am I overlooking a simpler approach?
> >
> > If I have to go this way, what sort of position score  code might
> work out
> > here?
> >
> > TIA
> > Sid
> >
> >
> > ---
> > Outgoing mail is certified Virus Free.
> > Checked by AVG anti-virus system (http://www.grisoft.com).
> > Version: 6.0.762 / Virus Database: 510 - Release Date: 9/13/2004
> >
>
>
>
>
>
>
>
>
>
>
> Yahoo! Groups Links
>







Yahoo! Groups Links

Hi G.,

Well isn't that what we'd all like?  ^^_^^

IMHO, very difficult to do.  You might have better luck if you
qualified your openings, however.  But you might find, for example,
that serious gap openings should be faded on the open, but then
reversed to trade in the direction of the gap on the close, and even
that kind of trading almost surely needs testing and qualification.
Not everyone has the stomach for this, and it hardly meets your goal
of a simple system that you don't have to worry about.

Considering all symbols at all times, I think I have seen studies
that indicate the Open/Close relationship is pretty darn close to
random.

As for advisory services ... if they could advise you profitably, why
would they?  They could make far more money acting on their own
knowledge.  They don't, because they haven't got anymore knowledge
than you do, so enough said on that.  Spend your money on something
worthwhile, rather than on someone who "has a system", but who is
selling it rather than trading it.

Any published successful system will soon be pretty much worthless,
so enough on that, too.

You are asking for a sort of modified holy grail.  Good luck.

Amibroker has all the capabilities you need to test any assertions
I've made, or find out the probabilities of your own market for
yourself.

Yuki

Thursday, September 21, 2006, 4:08:00 PM, you wrote:

GFN> Does anyone have any ideas on how difficult it is to predict the
GFN> position of the Open and Close relative to each other.  For example an
GFN> Open at 650 and a Close of 655
GFN> would be an up day, ignoring pervious market activity. The goal here
GFN> is a simple day trading system that I don't have to worry about while
GFN> at work, just buy or sell at Open and exit at Close.

GFN> Any ideas on the accuracy ratio of published systems or advisory
GFN> services in doing this?

GFN> GFN

To minimize uninformed ranting, Bob, you may want to check out this
website which has links to 136 technical papers in which MCS is
discussed/explored mostly with respect to VaR... and this is just the
tip of the iceberg :-)

http://www.gloriamundi.org/

--- In amibroker-ts@yahoogroups.com, "Bob Jagow" <bjagow@...> wrote:
>
> It's really a Greek thing, 714, but I think further ranting should await
> responses from a modicum of the other 713 quanttraders :)
>
> -----Original Message-----
> From: amibroker-ts@yahoogroups.com
> [mailto:amibroker-ts@yahoogroups.com]On Behalf Of quanttrader714
> Sent: Wednesday, September 20, 2006 11:49 AM
> To: amibroker-ts@yahoogroups.com
> Subject: [amibroker-ts] Re: trading system subtleties, level 1
>
>
> You say that as if MCS is somehow a questionable or fringe method.  It
> is not.  It is used in some capacity by probably every large financial
> institution in the world, and has tons of applications in other fields
> as well, some of which are listed here.
>
> http://en.wikipedia.org/wiki/Monte_Carlo_method
>
> --- In amibroker-ts@yahoogroups.com, "Bob Jagow" <bjagow@> wrote:
> >
> > So its success with nuclear fallout gives it credence?
> >
> > I'd limit Monte Carlo calcs to mathematical problems that I couldn't
> [or was
> > too lazy to] analyze.
> >
> > Bob
> >
> > -----Original Message-----
> > From: amibroker-ts@yahoogroups.com
> > [mailto:amibroker-ts@yahoogroups.com]On Behalf Of brpnw1
> > Sent: Monday, September 18, 2006 6:31 PM
> > To: amibroker-ts@yahoogroups.com
> > Subject: [amibroker-ts] Re: trading system subtleties, level 1
> >
> >
> > I am a big fan of Monte Carlo Simulation (MCS), so as much as
> > backtesting makes sense, I don't think it really means much unless
> > you followup your results with MCS to make sure your results are
> > real.
> >
> > One idea is to optimize your system against the total of all symbols
> > that you are interested in trading.
> >
> > Then, trade those settings against each individual symbol, one at a
> > time. Take each results set for each symbol, and check it against
> > MCS. The best performers will be the ones with the best MCS ratios.
> >
> > I am not a statistician, although my experience with backtesting and
> > MCS is that MCS is profoundly more reliable in its results. MCS was
> > originally developed to predict where nuclear fallout would end up,
> > on and around nuclear test ranges. It was highly accurate from the
> > start of its existence.
> >
> > I would LOVE to see automatic MCS integration with AB. If anyone has
> > a script for doing this within AB, or in conjunction with an Excel
> > doc that handles the MCS, that would be fantastic!! I already have
> > an Excel doc for handling the MCS. Just would like to automate the
> > ongoing system backtesting against MCS, that's all.
> >
> > Thanks,
> >
> > Brian
> >
> > --- In amibroker-ts@yahoogroups.com, Sidney Kaiser <s9kaiser@>
> > wrote:
> > >
> > > I was doing a little "simple", or so I thought, trading system
> > work this
> > > weekend and was reminded once again how it is not simple.
> > >
> > > System: StochRSI fixed trigger levels, appx buy 20 sell 80
> > > 	 uses a trend filter appx 25-45 SMA, only take trades
> > in the direction of
> > > the trend
> > > 	 trailing stop
> > > 	 long only, daily data
> > > 	 start date 1/2/2003
> > >
> > > Stock list:  I started with the One Chicago list of single stock
> > futures,
> > > appx 120 issues.
> > > 	 ran a coarse optimization of params to find which
> > stocks would trade well
> > > with StochRSI
> > > 	 created a new watch list with the issues that had >
> > 10% profit over the
> > > test period
> > > 	 using the old V4.4 backtester to quickly generate
> > stock by stock results,
> > > 58 stocks
> > >
> > > I followed that up with a final optimization of params on the 58
> > stock list
> > > then ran the default portfolio backtester.  Starting with $10k the
> > profit
> > > was $68.6k.
> > >
> > > I then changed the stock list slightly and ended up with 62
> > stocks, the
> > > portfolio BT profit was $32.7k.
> > >
> > > Changing the stock list selection to > 20% profit and re-testing
> > with 49
> > > stocks, the portfolio BT profit was $24.4k.
> > >
> > > At first I thought, this was weird, but then I remembered
> > something I first
> > > learned years ago when doing backtests with FastTools.  The
> > portfolio
> > > tester is choosing one stock at a time from the list of available
> > > stocks.  If I slightly change the composition of the list, a new,
> > unique,
> > > sequence of trades is generated with very different results.  This
> > is
> > > something that I need to keep in the forefront of my thinking when
> > I am
> > > evaluating backtest results.
> > >
> > > It is very difficult to evaluate the effects of changing the list
> > of stocks
> > > because of this sequencing of trades effect.
> > >
> > > At this point I am faced with doing a full blown position score
> > approach to
> > > the backtesting.  That is unfortunate because I have never been
> > very
> > > successful with devising useful position scoring routines.
> > >
> > > Am I overlooking a simpler approach?
> > >
> > > If I have to go this way, what sort of position score  code might
> > work out
> > > here?
> > >
> > > TIA
> > > Sid
> > >
> > >
> > > ---
> > > Outgoing mail is certified Virus Free.
> > > Checked by AVG anti-virus system (http://www.grisoft.com).
> > > Version: 6.0.762 / Virus Database: 510 - Release Date: 9/13/2004
> > >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> > Yahoo! Groups Links
> >
>
>
>
>
>
>
>
> Yahoo! Groups Links
>

Yuki,

Thank you for your detailed and insightful comments.

In response to what I think are some of your key comments:

1) "It is difficult"
I agree that it's very difficult. Yet there are services that claim
to have an 80% accuracy in forecasting the daily direction of the
market. There is even a company that sells a proprietary DLL that
supposedly gives the 'next bar" direction with significant accuracy.

2) "qualified your openings"

Yes, I've found this is very important. By creating a 5 x 5 decision
matrix on the opening a lot of standard trading systems can vastly
improve performance. I use the following elements 1) gap up (open
above high) 2) strong (above close by significant amount) 3) even
(within a range based on recent volitility 4) weak 5) gap down. The
other dimension is based on prior behavior

3) "> As for advisory services ... if they could advise you
profitably, why would they?"

This is the ultimate question...why if you have a system that can
return 200% per year or has a 9 to 1 Win Loss ratio why sell it?
Just start a hedge fund!  Perhaps there are conditions where you'd
need alternatives. Let's say you are a math student or engineer and
you stumble on something that works in the markets that is an
established method in your field but is so esoteric that it's never
been tried on the markets. For example I know some who are using
very advanced spectral analysis from physics on the markets and
claim to be doing well. So if you are a grad student you may not
have the resources to start a company. I guess then you need to
offer a service or sell something to get going. Perhaps the best
approach would be to offer the output of your method to an
established hedge fund to get started?

4) "Any published successful system will soon be pretty much
worthless"

I'm not sure about that. I know people who claim to have made
consistent, but modest, profits for several decades using moving
average crossovers. In my own research I can see that if you can
resolve the problem of false crossings, it will work. That one big
trend that happens every 6 or 10 months can make a ton of money if
you can keep the 5 to 8 false crossings from eating up all your
profits. Some folks think money management is the answer. I'm
working on using Neural Network models to weed out the false
crossings. MA stuff has been around for decades but still works or
shows that it can work with the above qualification.

Personally I think some things continue to work to some degree
because of two things:

A) They reflect an underlying reality like gravity. If markets go
down far enough there is a point where enough people will thinks the
price is a bargain and start to buy it.

B) The consciousness of investors has a chaotic element. Human
decisions are not based on objective evidence. System A may have a
different risk potential then system B. People will choose one over
the other based not on accuracy or net profit but on the experience
and amount of risk. Some people are averse to holding a loosing
position even for a week while the markets finishes looking for the
last bottom before a sustained up move and will ignore such systems.

  I think that is the short coming in the Efficient Market Hypothesis
(EMH). The assumption that everyone has the same information, and
knows the dependability of that information and will act on it
objectively doesn't match the quirks of human behavior.

You and I could be presented with a trading system. Let's say we
are. Out of sample testing and live paper trading shows a return on
account of 500% per year and a Win Loss ratio of 91%. I may say this
is impossible and reject every trusting or trading the system. You
on the other hand may say "let's test it", start trading it and make
a fortune.

5) "You are asking for a sort of modified holy grail"

I used to think so too....until recently. I've come upon something
that may not be the grail itself but like in the DeVinci Code points
to where it must be found.

Let me know if this interest you and I'll keep you posted on
developments.

Fredrick



--- In amibroker-ts@yahoogroups.com, Yuki Taga <yukitaga@...> wrote:
>
> Hi G.,
>
> Well isn't that what we'd all like?  ^^_^^
>
> IMHO, very difficult to do.  You might have better luck if you
> qualified your openings, however.  But you might find, for example,
> that serious gap openings should be faded on the open, but then
> reversed to trade in the direction of the gap on the close, and
even
> that kind of trading almost surely needs testing and qualification.
> Not everyone has the stomach for this, and it hardly meets your
goal
> of a simple system that you don't have to worry about.
>
> Considering all symbols at all times, I think I have seen studies
> that indicate the Open/Close relationship is pretty darn close to
> random.
>
> As for advisory services ... if they could advise you profitably,
why
> would they?  They could make far more money acting on their own
> knowledge.  They don't, because they haven't got anymore knowledge
> than you do, so enough said on that.  Spend your money on something
> worthwhile, rather than on someone who "has a system", but who is
> selling it rather than trading it.
>
> Any published successful system will soon be pretty much worthless,
> so enough on that, too.
>
> You are asking for a sort of modified holy grail.  Good luck.
>
> Amibroker has all the capabilities you need to test any assertions
> I've made, or find out the probabilities of your own market for
> yourself.
>
> Yuki
>
> Thursday, September 21, 2006, 4:08:00 PM, you wrote:
>
> GFN> Does anyone have any ideas on how difficult it is to predict
the
> GFN> position of the Open and Close relative to each other.  For
example an
> GFN> Open at 650 and a Close of 655
> GFN> would be an up day, ignoring pervious market activity. The
goal here
> GFN> is a simple day trading system that I don't have to worry
about while
> GFN> at work, just buy or sell at Open and exit at Close.
>
> GFN> Any ideas on the accuracy ratio of published systems or
advisory
> GFN> services in doing this?
>
> GFN> GFN
>

Hello Frederick,

I'm a bit late in replying but...

Re #1 below ( services forecasting daily market direction ) -
Would you happen to have any names or URL's?  I would be interested in
having a look anyway...

Re #5 below ( DaVinci code for the grail ) -
It interests me!    8 - )
I think it would be great if you could keep us posted on developments...

Thanks very much!

Steve

----- Original Message -----
From: "G. Fredrick Nowatzke" <nowatzke60130@...>
To: <amibroker-ts@yahoogroups.com>
Sent: Thursday, September 21, 2006 12:53 PM
Subject: [amibroker-ts] Re: Market Direction


> Yuki,
>
> Thank you for your detailed and insightful comments.
>
> In response to what I think are some of your key comments:
>
> 1) "It is difficult"
> I agree that it's very difficult. Yet there are services that claim
> to have an 80% accuracy in forecasting the daily direction of the
> market. There is even a company that sells a proprietary DLL that
> supposedly gives the 'next bar" direction with significant accuracy.
>
> 2) "qualified your openings"
>
> Yes, I've found this is very important. By creating a 5 x 5 decision
> matrix on the opening a lot of standard trading systems can vastly
> improve performance. I use the following elements 1) gap up (open
> above high) 2) strong (above close by significant amount) 3) even
> (within a range based on recent volitility 4) weak 5) gap down. The
> other dimension is based on prior behavior
>
> 3) "> As for advisory services ... if they could advise you
> profitably, why would they?"
>
> This is the ultimate question...why if you have a system that can
> return 200% per year or has a 9 to 1 Win Loss ratio why sell it?
> Just start a hedge fund!  Perhaps there are conditions where you'd
> need alternatives. Let's say you are a math student or engineer and
> you stumble on something that works in the markets that is an
> established method in your field but is so esoteric that it's never
> been tried on the markets. For example I know some who are using
> very advanced spectral analysis from physics on the markets and
> claim to be doing well. So if you are a grad student you may not
> have the resources to start a company. I guess then you need to
> offer a service or sell something to get going. Perhaps the best
> approach would be to offer the output of your method to an
> established hedge fund to get started?
>
> 4) "Any published successful system will soon be pretty much
> worthless"
>
> I'm not sure about that. I know people who claim to have made
> consistent, but modest, profits for several decades using moving
> average crossovers. In my own research I can see that if you can
> resolve the problem of false crossings, it will work. That one big
> trend that happens every 6 or 10 months can make a ton of money if
> you can keep the 5 to 8 false crossings from eating up all your
> profits. Some folks think money management is the answer. I'm
> working on using Neural Network models to weed out the false
> crossings. MA stuff has been around for decades but still works or
> shows that it can work with the above qualification.
>
> Personally I think some things continue to work to some degree
> because of two things:
>
> A) They reflect an underlying reality like gravity. If markets go
> down far enough there is a point where enough people will thinks the
> price is a bargain and start to buy it.
>
> B) The consciousness of investors has a chaotic element. Human
> decisions are not based on objective evidence. System A may have a
> different risk potential then system B. People will choose one over
> the other based not on accuracy or net profit but on the experience
> and amount of risk. Some people are averse to holding a loosing
> position even for a week while the markets finishes looking for the
> last bottom before a sustained up move and will ignore such systems.
>
> I think that is the short coming in the Efficient Market Hypothesis
> (EMH). The assumption that everyone has the same information, and
> knows the dependability of that information and will act on it
> objectively doesn't match the quirks of human behavior.
>
> You and I could be presented with a trading system. Let's say we
> are. Out of sample testing and live paper trading shows a return on
> account of 500% per year and a Win Loss ratio of 91%. I may say this
> is impossible and reject every trusting or trading the system. You
> on the other hand may say "let's test it", start trading it and make
> a fortune.
>
> 5) "You are asking for a sort of modified holy grail"
>
> I used to think so too....until recently. I've come upon something
> that may not be the grail itself but like in the DeVinci Code points
> to where it must be found.
>
> Let me know if this interest you and I'll keep you posted on
> developments.
>
> Fredrick
>
>
>
> --- In amibroker-ts@yahoogroups.com, Yuki Taga <yukitaga@...> wrote:
>>
>> Hi G.,
>>
>> Well isn't that what we'd all like?  ^^_^^
>>
>> IMHO, very difficult to do.  You might have better luck if you
>> qualified your openings, however.  But you might find, for example,
>> that serious gap openings should be faded on the open, but then
>> reversed to trade in the direction of the gap on the close, and
> even
>> that kind of trading almost surely needs testing and qualification.
>> Not everyone has the stomach for this, and it hardly meets your
> goal
>> of a simple system that you don't have to worry about.
>>
>> Considering all symbols at all times, I think I have seen studies
>> that indicate the Open/Close relationship is pretty darn close to
>> random.
>>
>> As for advisory services ... if they could advise you profitably,
> why
>> would they?  They could make far more money acting on their own
>> knowledge.  They don't, because they haven't got anymore knowledge
>> than you do, so enough said on that.  Spend your money on something
>> worthwhile, rather than on someone who "has a system", but who is
>> selling it rather than trading it.
>>
>> Any published successful system will soon be pretty much worthless,
>> so enough on that, too.
>>
>> You are asking for a sort of modified holy grail.  Good luck.
>>
>> Amibroker has all the capabilities you need to test any assertions
>> I've made, or find out the probabilities of your own market for
>> yourself.
>>
>> Yuki
>>
>> Thursday, September 21, 2006, 4:08:00 PM, you wrote:
>>
>> GFN> Does anyone have any ideas on how difficult it is to predict
> the
>> GFN> position of the Open and Close relative to each other.  For
> example an
>> GFN> Open at 650 and a Close of 655
>> GFN> would be an up day, ignoring pervious market activity. The
> goal here
>> GFN> is a simple day trading system that I don't have to worry
> about while
>> GFN> at work, just buy or sell at Open and exit at Close.
>>
>> GFN> Any ideas on the accuracy ratio of published systems or
> advisory
>> GFN> services in doing this?
>>
>> GFN> GFN
>>
>
>
>
>
>
>
>
> Yahoo! Groups Links
>
>
>
>
>
>
>
>
>
>
>

I am looking for some help regarding automatic order generated by AB.
My  problem is that I have redundant orders that are sent to IB based
on signals. I don't know how to handle this problem. Any ideas?

I wanted to comment on this too.

Let's say that there was a system that did very well that was discovered by
accident and somone posted it on a blog.

How many people would actually follow it to the point of making it
worthless?

How many people would follow it correctly, even?

There's even simple trading systems out there that people royally screw up
because of their emotions.

Even look at poker. There's basic concepts that you can use to beat many low
limit games and still many people screw up at that.

Don't discount published systems until you test them.

-Mike



>
> 4) "Any published successful system will soon be pretty much
> worthless"
>
> I'm not sure about that. I know people who claim to have made
> consistent, but modest, profits for several decades using moving
> average crossovers. In my own research I can see that if you can
> resolve the problem of false crossings, it will work. That one big
> trend that happens every 6 or 10 months can make a ton of money if
> you can keep the 5 to 8 false crossings from eating up all your
> profits. Some folks think money management is the answer. I'm
> working on using Neural Network models to weed out the false
> crossings. MA stuff has been around for decades but still works or
> shows that it can work with the above qualification.
>
>


[Non-text portions of this message have been removed]

Hello,

I try to put an automated trading system with Amibroker+IB plugin
database+IBController linked to TWS.

Actually all is ok. Problem is just i need to have in real
time "Buy/Sell/Short/Cover" signal from 100 tickers (stocks) who are
update in real time via IB plugin for realtime database.
Check must be done every minute, synchronised with the realtime data
(1min chart).

To be clear: 1 minute chart and trading intervalle, 100 stock real
time. I need to send order to IB controller as fast as possible
after the end of a new minute.
So when a minute is finish:
- a new close is updated in the database for the 100 real time
tickers
- a new scan is launched ("Run every: 1min")
- buy/sell/cover/short signal are generated and in the same way
order are send to IBController


My problem is the time delay betwenn the new close and the order
sent. To resume :
- TWS Data -> Amibroker (via IB Plug ins) : around 0.5sec delay
(because 100 realtime tickers it seems !)
- Scan the 100 tickers: Too long, 3 sec with to 30 days data / 1 min
bars !
- Send the order to IBController and IB Controller to TWS: very fast
(less than 0.3 sec i think).

Problem is more the time between the new close and the order is
receive in TWS, more the slippage is important... and it can ruins
all the system for sure.


My questions are :
- for the scan : when i select just 1 day range or 30 last
quotations, it takes the same time than 30 day range... but i only
need to scan the last 30 quotations.
It is really a big problem for real time trading. How can i have
faster scan on only the last quotations without scan through all the
data.
- maybe there is another way to proced for realtime trading 1min
intervalles ? put 100 indactors layout (one for each tickers ?)


Thank you,
Mich.

--- In amibroker-ts@yahoogroups.com, "Kal Parekh" <parekk@...> wrote:
>
> Hi All,
>
> Thank you for time and support.
>
> Quick ques: Can anyone point to AB code that allows me simulate exit
> after n-bars if a position is unprofitable? i.e, exit trade after 10
> days is the trade is unprofitable.
>
> Again, thank you for your time and support.
>
> KP
>


Try this:

Replace normal SELL with SELLA

SELLB = iif (ref(Buy,-10) and C < BUYPRICE, 1,0);

SELL = SELLA or SELLB;



Tomas

Hi Fredrick,

In line ...

Friday, September 22, 2006, 1:53:58 AM, you wrote:

GFN> 1) "It is difficult"
GFN> I agree that it's very difficult. Yet there are services that claim
GFN> to have an 80% accuracy in forecasting the daily direction of the
GFN> market. There is even a company that sells a proprietary DLL that
GFN> supposedly gives the 'next bar" direction with significant accuracy.

Anyone can claim anything, of course.  Moreover, one can win 80
percent of one's trades ... and merrily go bankrupt in the process of
doing so.  Well, okay, maybe it wouldn't be so merry after all, but
there are other metrics that are essential to know and understand.

Think what you (as a shrewd person) would demand from a vendor before
you would: A) pay that vendor to give you advice, and B) speculate
putting up more than 100 percent of the risk (you had to pay the
vendor first) based on that vendor's advice. Would you accept a black
box dll?  Backtest claims that you could not verify?  I don't think
so.

Cigarette manufacturers still claim that their products are not
linked to cancer, emphysema, or a host of other nasty things.

The point here is that there are a lot of people who are *not*
shrewd, leaving the field open for hucksters who claim lots of things
that have no basis in reality.

One wonders why these people are in business.  The reason they are is
that "a sucker born every minute" grossly understates the reality.

GFN> 2) "qualified your openings"

GFN> Yes, I've found this is very important. By creating a 5 x 5 decision
GFN> matrix on the opening a lot of standard trading systems can vastly
GFN> improve performance. I use the following elements 1) gap up (open
GFN> above high) 2) strong (above close by significant amount) 3) even
GFN> (within a range based on recent volitility 4) weak 5) gap down. The
GFN> other dimension is based on prior behavior

Certainly the way I would begin to approach the search for an edge.
But be careful about degrees of freedom.

GFN> 3) "> As for advisory services ... if they could advise you
GFN> profitably, why would they?"

GFN> This is the ultimate question...why if you have a system that can
GFN> return 200% per year or has a 9 to 1 Win Loss ratio why sell it?
GFN> Just start a hedge fund!  Perhaps there are conditions where you'd
GFN> need alternatives. Let's say you are a math student or engineer and
GFN> you stumble on something that works in the markets that is an
GFN> established method in your field but is so esoteric that it's never
GFN> been tried on the markets. For example I know some who are using
GFN> very advanced spectral analysis from physics on the markets and
GFN> claim to be doing well. So if you are a grad student you may not
GFN> have the resources to start a company. I guess then you need to
GFN> offer a service or sell something to get going. Perhaps the best
GFN> approach would be to offer the output of your method to an
GFN> established hedge fund to get started?

This basically goes back to item number 1.  I don't know why, but
every time this conversation comes up (research it), it's always the
graduate student in engineering or math that has the system.  Every
time. It seems to never fail.  ^^_^^  No kidding.  It's always a grad
student.  Undergrads never develop successful systems it seems, so
never buy one from one.

As for said grad student not having resources, she would have to be
clever enough to find a system, but incredibly dull about finance in
general.  That is to say, a proven, demonstrable edge, something
significant that is long-lasting, robust, avoids major disasters, and
in general treats the market as an ATM machine will only require the
developer to approach the right people.  There are a lot of shrewd
people with a lot of money. Show any number of them that what you
have is something special, and you are going to get funded in a big
hurry.  I guarantee it.  But don't expect to try and get funding for
a black box, of course.

The ultimate problem is ... to prove the edge isn't a fluke or a
curve fit, you almost *have* to reveal the complete method to
someone, so they can duplicate your results.  You could,
alternatively, demonstrate the output in real time going forward, but
how many years would the recipient want to follow it before deciding
you probably didn't just curve fit something? No, you are almost
surely going to have to reveal the system -- before you get paid for
it, which means that there is the potential to give away your
proprietary edge before you've been compensated for it. So on that
one basis alone, I find some really weak support for the idea of
selling black-box output.  But that objection only works if no one is
honest.  I don't accept that proposition, either, although in any
business in which there is a lot of money lying all over the place,
keep track of where your purse is at all times.

The only way a black box system can be sold, I suspect, is that the
buyers have to not be very shrewd.  But hey, since the pool of not
shrewd is seemingly much, much larger than the pool of shrewd, maybe
I'm in the wrong vector of this business!

^^_^^

GFN> 4) "Any published successful system will soon be pretty much
GFN> worthless"

GFN> I'm not sure about that. I know people who claim to have made
GFN> consistent, but modest, profits for several decades using moving
GFN> average crossovers. In my own research I can see that if you can
GFN> resolve the problem of false crossings, it will work. That one big
GFN> trend that happens every 6 or 10 months can make a ton of money if
GFN> you can keep the 5 to 8 false crossings from eating up all your
GFN> profits. Some folks think money management is the answer. I'm
GFN> working on using Neural Network models to weed out the false
GFN> crossings. MA stuff has been around for decades but still works or
GFN> shows that it can work with the above qualification.

What a huge qualification, however.  Isn't that something like saying
"published systems will work if you can just get rid of the losers"?
Well, I suspect that's indeed true.  ^_-

GFN> Personally I think some things continue to work to some degree
GFN> because of two things:

GFN> A) They reflect an underlying reality like gravity. If markets go
GFN> down far enough there is a point where enough people will thinks the
GFN> price is a bargain and start to buy it.

GFN> B) The consciousness of investors has a chaotic element. Human
GFN> decisions are not based on objective evidence. System A may have a
GFN> different risk potential then system B. People will choose one over
GFN> the other based not on accuracy or net profit but on the experience
GFN> and amount of risk. Some people are averse to holding a loosing
GFN> position even for a week while the markets finishes looking for the
GFN> last bottom before a sustained up move and will ignore such systems.

GFN>  I think that is the short coming in the Efficient Market Hypothesis
GFN> (EMH). The assumption that everyone has the same information, and
GFN> knows the dependability of that information and will act on it
GFN> objectively doesn't match the quirks of human behavior.

I have a similar, if slightly different view.  First, the EMH is
flawed, for not only the reasons you cited just above, but also
because there is a tremendous amount of subjective, emotional trade
that takes place in every market every day.  No one can *possibly*
have perfect information, because everyone cannot know the emotional
output that is going to come from any number of other players, at any
time.  All we can know is that it is there, and will manifest itself
intermittently and without warning.

GFN> You and I could be presented with a trading system. Let's say we
GFN> are. Out of sample testing and live paper trading shows a return on
GFN> account of 500% per year and a Win Loss ratio of 91%. I may say this
GFN> is impossible and reject every trusting or trading the system. You
GFN> on the other hand may say "let's test it", start trading it and make
GFN> a fortune.

Of course I'm going to say "let's test it".  And so are you, unless
the source of the claim is so unreliable that she might have been the
source of all pre-invasion intelligence about Iraq for the US
government. ^^_^^ (Who says Yuki doesn't have a sense of humor?)

But seriously ... you are going to test something like that.  And if
it meets your very rigorous review, you are going to trade it, too.
It's the very vigorous review that is where the rub is.

But the simple and obvious problem with published systems is: They
get arbitraged to death.  First, we must assume that the system in
question is making a good enough return to be of envy, rather than
something that scalps a below-money market rate of return each year.
Then we must realize that, if good enough, and published, it's going
to get famous very, very quickly.  Today, it would be known by the
world, literally, in days or weeks.  And we know, instinctively, that
*any* system has a capacity, beyond which it simply cannot continue
to work.  It's basically the same reason Ponzi schemes fail, although
their math problem is perhaps much easier to understand.  But "free
money" attracts flies so quickly that you show people how to get
some, you will think you are living in Western Australia during the
summer. (Hey, I've been there ...) Remember, one of the two primary
emotions that drives the whole show is called greed. Sure, everyone
is going to have a slightly different idea of what is good enough to
trade, but we are talking about two things here: systems good enough
to buy or sell, and published systems that are also worth trading. I
don't think anyone with the former is going to risk publishing it.

But I'd love to read and understand one.  ^_^

GFN> 5) "You are asking for a sort of modified holy grail"

GFN> I used to think so too....until recently. I've come upon something
GFN> that may not be the grail itself but like in the DeVinci Code points
GFN> to where it must be found.

GFN> Let me know if this interest you and I'll keep you posted on
GFN> developments.

I'm always here ... and always at my e-mail address.  ^_^

And I'm always very, very skeptical until I see *all* the inner
workings.

Yuki

GFN> --- In amibroker-ts@yahoogroups.com, Yuki Taga <yukitaga@...> wrote:
>>
>> Hi G.,
>>
>> Well isn't that what we'd all like?  ^^_^^
>>
>> IMHO, very difficult to do.  You might have better luck if you
>> qualified your openings, however.  But you might find, for example,
>> that serious gap openings should be faded on the open, but then
>> reversed to trade in the direction of the gap on the close, and
GFN> even
>> that kind of trading almost surely needs testing and qualification.
>> Not everyone has the stomach for this, and it hardly meets your
GFN> goal
>> of a simple system that you don't have to worry about.
>>
>> Considering all symbols at all times, I think I have seen studies
>> that indicate the Open/Close relationship is pretty darn close to
>> random.
>>
>> As for advisory services ... if they could advise you profitably,
GFN> why
>> would they?  They could make far more money acting on their own
>> knowledge.  They don't, because they haven't got anymore knowledge
>> than you do, so enough said on that.  Spend your money on something
>> worthwhile, rather than on someone who "has a system", but who is
>> selling it rather than trading it.
>>
>> Any published successful system will soon be pretty much worthless,
>> so enough on that, too.
>>
>> You are asking for a sort of modified holy grail.  Good luck.
>>
>> Amibroker has all the capabilities you need to test any assertions
>> I've made, or find out the probabilities of your own market for
>> yourself.
>>
>> Yuki
>>
>> Thursday, September 21, 2006, 4:08:00 PM, you wrote:
>>
>> GFN> Does anyone have any ideas on how difficult it is to predict
GFN> the
>> GFN> position of the Open and Close relative to each other.  For
GFN> example an
>> GFN> Open at 650 and a Close of 655
>> GFN> would be an up day, ignoring pervious market activity. The
GFN> goal here
>> GFN> is a simple day trading system that I don't have to worry
GFN> about while
>> GFN> at work, just buy or sell at Open and exit at Close.
>>
>> GFN> Any ideas on the accuracy ratio of published systems or
GFN> advisory
>> GFN> services in doing this?
>>
>> GFN> GFN
>>






Best,

Yuki

Hello,

See the example code at http://www.amibroker.com/at/

You need to use STATIC variables to prevent double orders (so you know you
already placed the order).
because PORTFOLIO reporting in TWS is a bit delayed.

if( NOT StaticVarGet("OrderAlreadyPlacedFor" + Name() ) )
{
    StaticVarSet("OrderAlreadyPlacedFor"+Name(), 1 );

    ibc.PlaceOrder(....
}



Best regards,
Tomasz Janeczko
amibroker.com
----- Original Message -----
From: "teuffy" <teuffy@...>
To: <amibroker-ts@yahoogroups.com>
Sent: Thursday, September 21, 2006 6:34 PM
Subject: [amibroker-ts] Experience with BrokerIB


>I am looking for some help regarding automatic order generated by AB.
> My  problem is that I have redundant orders that are sent to IB based
> on signals. I don't know how to handle this problem. Any ideas?
>
>
>
>
>
>
>
>
>
> Yahoo! Groups Links
>
>
>
>
>
>
>
>
>
>
>

From my brief tangle with AB u must get yre hands dirty with the
BrokerIb commands.

You then must haveloops that check the status of yre trade and see if
it is open or closed. The amount of data available isn't huge from
what i see but u need to slowly build up knowledge of the commands in
AB's broker interface language .


regards
moz


--- In amibroker-ts@yahoogroups.com, "teuffy" <teuffy@...> wrote:
>
> I am looking for some help regarding automatic order generated by AB.
> My  problem is that I have redundant orders that are sent to IB based
> on signals. I don't know how to handle this problem. Any ideas?
>

HI,
I have stock data in my NASDAQ AB database that has ceased trading (
no more OHCL data after say eg. may 2006 ). When  back testing
sometimes my trading systems do not exit these stocks leaving them in
an OPEN POSITION till the end of period.
The problem is that this stock would have been exited in MAY06 and
another trade taken with available equity .

I can solve this problem by ignoring this stock but it would be better
if i could exit this STOCK if I could detect it has ceased trading .

Does anyone know how I could code this or ideas ?

Regards
Moztrader

Hi ,
u can use the follwoing in yre AA code




stock1 = IIf(StrFind(Name() , "goog") > 410,1,0 ) ;
stock2 = IIf(StrFind(Name() , "MSFT") > 120,1,0 ) ;
stock3 = <more stocks here > ;

FILTER = stock1 or stock2 or stock3 ;

// use addcolumn() to define what parameters u want to display ;
AddColumn( Close, "Close", 1.2 );
AddColumn( MACD(), "MACD", 1.4 , IIf( MACD() > 0, colorGreen, colorRed
) );
AddTextColumn( FullName(), "Full name", 77 , colorDefault, IIf( Close
< 10, colorLightBlue, colorDefault ) );

hope this helps
moz


--- In amibroker-ts@yahoogroups.com, "Dorothy" <dora_phuki@...> wrote:
>
> i am a newbie...can anyone help? can i code in automatic analysis
> where i can just check levels of certain stock specified in the scan
> (hard coded) e.g. GOOG > 410 or MSFT greater > 120 in eod data and
> generate an alert?
>
> thanks
>

Yuki,



I truly agree with all your points, but have one piece of math to be
considered because I always wondered why anyone would sell their winning
systems:



10,000 customers * $100/month = $12,000,000 per year.that's twelve
million dollars (1.264 billion yen).

Zero drawdown.

Zero risk.

Zero commissions.

Minimal expenses.



I think this why they would sell their system even if it IS a winner,
but especially if it's not ;-)

--

Terry

-----Original Message-----
From: amibroker-ts@yahoogroups.com [mailto:amibroker-ts@yahoogroups.com]
On Behalf Of Yuki Taga
Sent: Thursday, September 21, 2006 22:21
To: G. Fredrick Nowatzke
Subject: Re: [amibroker-ts] Re: Market Direction



Hi Fredrick,

In line ...

Friday, September 22, 2006, 1:53:58 AM, you wrote:

GFN> 1) "It is difficult"
GFN> I agree that it's very difficult. Yet there are services that claim

GFN> to have an 80% accuracy in forecasting the daily direction of the
GFN> market. There is even a company that sells a proprietary DLL that
GFN> supposedly gives the 'next bar" direction with significant
accuracy.

Anyone can claim anything, of course. Moreover, one can win 80
percent of one's trades ... and merrily go bankrupt in the process of
doing so. Well, okay, maybe it wouldn't be so merry after all, but
there are other metrics that are essential to know and understand.

Think what you (as a shrewd person) would demand from a vendor before
you would: A) pay that vendor to give you advice, and B) speculate
putting up more than 100 percent of the risk (you had to pay the
vendor first) based on that vendor's advice. Would you accept a black
box dll? Backtest claims that you could not verify? I don't think
so.

Cigarette manufacturers still claim that their products are not
linked to cancer, emphysema, or a host of other nasty things.

The point here is that there are a lot of people who are *not*
shrewd, leaving the field open for hucksters who claim lots of things
that have no basis in reality.

One wonders why these people are in business. The reason they are is
that "a sucker born every minute" grossly understates the reality.

GFN> 2) "qualified your openings"

GFN> Yes, I've found this is very important. By creating a 5 x 5
decision
GFN> matrix on the opening a lot of standard trading systems can vastly
GFN> improve performance. I use the following elements 1) gap up (open
GFN> above high) 2) strong (above close by significant amount) 3) even
GFN> (within a range based on recent volitility 4) weak 5) gap down. The

GFN> other dimension is based on prior behavior

Certainly the way I would begin to approach the search for an edge.
But be careful about degrees of freedom.

GFN> 3) "> As for advisory services ... if they could advise you
GFN> profitably, why would they?"

GFN> This is the ultimate question...why if you have a system that can
GFN> return 200% per year or has a 9 to 1 Win Loss ratio why sell it?
GFN> Just start a hedge fund! Perhaps there are conditions where you'd
GFN> need alternatives. Let's say you are a math student or engineer and

GFN> you stumble on something that works in the markets that is an
GFN> established method in your field but is so esoteric that it's never

GFN> been tried on the markets. For example I know some who are using
GFN> very advanced spectral analysis from physics on the markets and
GFN> claim to be doing well. So if you are a grad student you may not
GFN> have the resources to start a company. I guess then you need to
GFN> offer a service or sell something to get going. Perhaps the best
GFN> approach would be to offer the output of your method to an
GFN> established hedge fund to get started?

This basically goes back to item number 1. I don't know why, but
every time this conversation comes up (research it), it's always the
graduate student in engineering or math that has the system. Every
time. It seems to never fail. ^^_^^ No kidding. It's always a grad
student. Undergrads never develop successful systems it seems, so
never buy one from one.

As for said grad student not having resources, she would have to be
clever enough to find a system, but incredibly dull about finance in
general. That is to say, a proven, demonstrable edge, something
significant that is long-lasting, robust, avoids major disasters, and
in general treats the market as an ATM machine will only require the
developer to approach the right people. There are a lot of shrewd
people with a lot of money. Show any number of them that what you
have is something special, and you are going to get funded in a big
hurry. I guarantee it. But don't expect to try and get funding for
a black box, of course.

The ultimate problem is ... to prove the edge isn't a fluke or a
curve fit, you almost *have* to reveal the complete method to
someone, so they can duplicate your results. You could,
alternatively, demonstrate the output in real time going forward, but
how many years would the recipient want to follow it before deciding
you probably didn't just curve fit something? No, you are almost
surely going to have to reveal the system -- before you get paid for
it, which means that there is the potential to give away your
proprietary edge before you've been compensated for it. So on that
one basis alone, I find some really weak support for the idea of
selling black-box output. But that objection only works if no one is
honest. I don't accept that proposition, either, although in any
business in which there is a lot of money lying all over the place,
keep track of where your purse is at all times.

The only way a black box system can be sold, I suspect, is that the
buyers have to not be very shrewd. But hey, since the pool of not
shrewd is seemingly much, much larger than the pool of shrewd, maybe
I'm in the wrong vector of this business!

^^_^^

GFN> 4) "Any published successful system will soon be pretty much
GFN> worthless"

GFN> I'm not sure about that. I know people who claim to have made
GFN> consistent, but modest, profits for several decades using moving
GFN> average crossovers. In my own research I can see that if you can
GFN> resolve the problem of false crossings, it will work. That one big
GFN> trend that happens every 6 or 10 months can make a ton of money if
GFN> you can keep the 5 to 8 false crossings from eating up all your
GFN> profits. Some folks think money management is the answer. I'm
GFN> working on using Neural Network models to weed out the false
GFN> crossings. MA stuff has been around for decades but still works or
GFN> shows that it can work with the above qualification.

What a huge qualification, however. Isn't that something like saying
"published systems will work if you can just get rid of the losers"?
Well, I suspect that's indeed true. ^_-

GFN> Personally I think some things continue to work to some degree
GFN> because of two things:

GFN> A) They reflect an underlying reality like gravity. If markets go
GFN> down far enough there is a point where enough people will thinks
the
GFN> price is a bargain and start to buy it.

GFN> B) The consciousness of investors has a chaotic element. Human
GFN> decisions are not based on objective evidence. System A may have a
GFN> different risk potential then system B. People will choose one over

GFN> the other based not on accuracy or net profit but on the experience

GFN> and amount of risk. Some people are averse to holding a loosing
GFN> position even for a week while the markets finishes looking for the

GFN> last bottom before a sustained up move and will ignore such
systems.

GFN> I think that is the short coming in the Efficient Market Hypothesis

GFN> (EMH). The assumption that everyone has the same information, and
GFN> knows the dependability of that information and will act on it
GFN> objectively doesn't match the quirks of human behavior.

I have a similar, if slightly different view. First, the EMH is
flawed, for not only the reasons you cited just above, but also
because there is a tremendous amount of subjective, emotional trade
that takes place in every market every day. No one can *possibly*
have perfect information, because everyone cannot know the emotional
output that is going to come from any number of other players, at any
time. All we can know is that it is there, and will manifest itself
intermittently and without warning.

GFN> You and I could be presented with a trading system. Let's say we
GFN> are. Out of sample testing and live paper trading shows a return on

GFN> account of 500% per year and a Win Loss ratio of 91%. I may say
this
GFN> is impossible and reject every trusting or trading the system. You
GFN> on the other hand may say "let's test it", start trading it and
make
GFN> a fortune.

Of course I'm going to say "let's test it". And so are you, unless
the source of the claim is so unreliable that she might have been the
source of all pre-invasion intelligence about Iraq for the US
government. ^^_^^ (Who says Yuki doesn't have a sense of humor?)

But seriously ... you are going to test something like that. And if
it meets your very rigorous review, you are going to trade it, too.
It's the very vigorous review that is where the rub is.

But the simple and obvious problem with published systems is: They
get arbitraged to death. First, we must assume that the system in
question is making a good enough return to be of envy, rather than
something that scalps a below-money market rate of return each year.
Then we must realize that, if good enough, and published, it's going
to get famous very, very quickly. Today, it would be known by the
world, literally, in days or weeks. And we know, instinctively, that
*any* system has a capacity, beyond which it simply cannot continue
to work. It's basically the same reason Ponzi schemes fail, although
their math problem is perhaps much easier to understand. But "free
money" attracts flies so quickly that you show people how to get
some, you will think you are living in Western Australia during the
summer. (Hey, I've been there ...) Remember, one of the two primary
emotions that drives the whole show is called greed. Sure, everyone
is going to have a slightly different idea of what is good enough to
trade, but we are talking about two things here: systems good enough
to buy or sell, and published systems that are also worth trading. I
don't think anyone with the former is going to risk publishing it.

But I'd love to read and understand one. ^_^

GFN> 5) "You are asking for a sort of modified holy grail"

GFN> I used to think so too....until recently. I've come upon something
GFN> that may not be the grail itself but like in the DeVinci Code
points
GFN> to where it must be found.

GFN> Let me know if this interest you and I'll keep you posted on
GFN> developments.

I'm always here ... and always at my e-mail address. ^_^

And I'm always very, very skeptical until I see *all* the inner
workings.

Yuki

GFN> --- In amibroker-ts@ <mailto:amibroker-ts%40yahoogroups.com>
yahoogroups.com, Yuki Taga <yukitaga@...> wrote:
>>
>> Hi G.,
>>
>> Well isn't that what we'd all like? ^^_^^
>>
>> IMHO, very difficult to do. You might have better luck if you
>> qualified your openings, however. But you might find, for example,
>> that serious gap openings should be faded on the open, but then
>> reversed to trade in the direction of the gap on the close, and
GFN> even
>> that kind of trading almost surely needs testing and qualification.
>> Not everyone has the stomach for this, and it hardly meets your
GFN> goal
>> of a simple system that you don't have to worry about.
>>
>> Considering all symbols at all times, I think I have seen studies
>> that indicate the Open/Close relationship is pretty darn close to
>> random.
>>
>> As for advisory services ... if they could advise you profitably,
GFN> why
>> would they? They could make far more money acting on their own
>> knowledge. They don't, because they haven't got anymore knowledge
>> than you do, so enough said on that. Spend your money on something
>> worthwhile, rather than on someone who "has a system", but who is
>> selling it rather than trading it.
>>
>> Any published successful system will soon be pretty much worthless,
>> so enough on that, too.
>>
>> You are asking for a sort of modified holy grail. Good luck.
>>
>> Amibroker has all the capabilities you need to test any assertions
>> I've made, or find out the probabilities of your own market for
>> yourself.
>>
>> Yuki
>>
>> Thursday, September 21, 2006, 4:08:00 PM, you wrote:
>>
>> GFN> Does anyone have any ideas on how difficult it is to predict
GFN> the
>> GFN> position of the Open and Close relative to each other. For
GFN> example an
>> GFN> Open at 650 and a Close of 655
>> GFN> would be an up day, ignoring pervious market activity. The
GFN> goal here
>> GFN> is a simple day trading system that I don't have to worry
GFN> about while
>> GFN> at work, just buy or sell at Open and exit at Close.
>>
>> GFN> Any ideas on the accuracy ratio of published systems or
GFN> advisory
>> GFN> services in doing this?
>>
>> GFN> GFN
>>

Best,

Yuki





[Non-text portions of this message have been removed]

simple enough - i am interested. keep me posted on the developments.


--- In amibroker-ts@yahoogroups.com, "G. Fredrick Nowatzke"
<nowatzke60130@...> wrote:
>
> Yuki,
>
> Thank you for your detailed and insightful comments.
>
> In response to what I think are some of your key comments:
>
> 1) "It is difficult"
> I agree that it's very difficult. Yet there are services that claim
> to have an 80% accuracy in forecasting the daily direction of the
> market. There is even a company that sells a proprietary DLL that
> supposedly gives the 'next bar" direction with significant accuracy.
>
> 2) "qualified your openings"
>
> Yes, I've found this is very important. By creating a 5 x 5 decision
> matrix on the opening a lot of standard trading systems can vastly
> improve performance. I use the following elements 1) gap up (open
> above high) 2) strong (above close by significant amount) 3) even
> (within a range based on recent volitility 4) weak 5) gap down. The
> other dimension is based on prior behavior
>
> 3) "> As for advisory services ... if they could advise you
> profitably, why would they?"
>
> This is the ultimate question...why if you have a system that can
> return 200% per year or has a 9 to 1 Win Loss ratio why sell it?
> Just start a hedge fund!  Perhaps there are conditions where you'd
> need alternatives. Let's say you are a math student or engineer and
> you stumble on something that works in the markets that is an
> established method in your field but is so esoteric that it's never
> been tried on the markets. For example I know some who are using
> very advanced spectral analysis from physics on the markets and
> claim to be doing well. So if you are a grad student you may not
> have the resources to start a company. I guess then you need to
> offer a service or sell something to get going. Perhaps the best
> approach would be to offer the output of your method to an
> established hedge fund to get started?
>
> 4) "Any published successful system will soon be pretty much
> worthless"
>
> I'm not sure about that. I know people who claim to have made
> consistent, but modest, profits for several decades using moving
> average crossovers. In my own research I can see that if you can
> resolve the problem of false crossings, it will work. That one big
> trend that happens every 6 or 10 months can make a ton of money if
> you can keep the 5 to 8 false crossings from eating up all your
> profits. Some folks think money management is the answer. I'm
> working on using Neural Network models to weed out the false
> crossings. MA stuff has been around for decades but still works or
> shows that it can work with the above qualification.
>
> Personally I think some things continue to work to some degree
> because of two things:
>
> A) They reflect an underlying reality like gravity. If markets go
> down far enough there is a point where enough people will thinks the
> price is a bargain and start to buy it.
>
> B) The consciousness of investors has a chaotic element. Human
> decisions are not based on objective evidence. System A may have a
> different risk potential then system B. People will choose one over
> the other based not on accuracy or net profit but on the experience
> and amount of risk. Some people are averse to holding a loosing
> position even for a week while the markets finishes looking for the
> last bottom before a sustained up move and will ignore such systems.
>
>  I think that is the short coming in the Efficient Market Hypothesis
> (EMH). The assumption that everyone has the same information, and
> knows the dependability of that information and will act on it
> objectively doesn't match the quirks of human behavior.
>
> You and I could be presented with a trading system. Let's say we
> are. Out of sample testing and live paper trading shows a return on
> account of 500% per year and a Win Loss ratio of 91%. I may say this
> is impossible and reject every trusting or trading the system. You
> on the other hand may say "let's test it", start trading it and make
> a fortune.
>
> 5) "You are asking for a sort of modified holy grail"
>
> I used to think so too....until recently. I've come upon something
> that may not be the grail itself but like in the DeVinci Code points
> to where it must be found.
>
> Let me know if this interest you and I'll keep you posted on
> developments.
>
> Fredrick
>
>
>
> --- In amibroker-ts@yahoogroups.com, Yuki Taga <yukitaga@> wrote:
> >
> > Hi G.,
> >
> > Well isn't that what we'd all like?  ^^_^^
> >
> > IMHO, very difficult to do.  You might have better luck if you
> > qualified your openings, however.  But you might find, for example,
> > that serious gap openings should be faded on the open, but then
> > reversed to trade in the direction of the gap on the close, and
> even
> > that kind of trading almost surely needs testing and qualification.
> > Not everyone has the stomach for this, and it hardly meets your
> goal
> > of a simple system that you don't have to worry about.
> >
> > Considering all symbols at all times, I think I have seen studies
> > that indicate the Open/Close relationship is pretty darn close to
> > random.
> >
> > As for advisory services ... if they could advise you profitably,
> why
> > would they?  They could make far more money acting on their own
> > knowledge.  They don't, because they haven't got anymore knowledge
> > than you do, so enough said on that.  Spend your money on something
> > worthwhile, rather than on someone who "has a system", but who is
> > selling it rather than trading it.
> >
> > Any published successful system will soon be pretty much worthless,
> > so enough on that, too.
> >
> > You are asking for a sort of modified holy grail.  Good luck.
> >
> > Amibroker has all the capabilities you need to test any assertions
> > I've made, or find out the probabilities of your own market for
> > yourself.
> >
> > Yuki
> >
> > Thursday, September 21, 2006, 4:08:00 PM, you wrote:
> >
> > GFN> Does anyone have any ideas on how difficult it is to predict
> the
> > GFN> position of the Open and Close relative to each other.  For
> example an
> > GFN> Open at 650 and a Close of 655
> > GFN> would be an up day, ignoring pervious market activity. The
> goal here
> > GFN> is a simple day trading system that I don't have to worry
> about while
> > GFN> at work, just buy or sell at Open and exit at Close.
> >
> > GFN> Any ideas on the accuracy ratio of published systems or
> advisory
> > GFN> services in doing this?
> >
> > GFN> GFN
> >
>

There may be a way to do what you want to do, but I just delete the
symbol altogether.  Your system should not stop working just because
one symbol was removed.

intermilan04

--- In amibroker-ts@yahoogroups.com, "moztrader" <das300@...> wrote:
>
> HI,
> I have stock data in my NASDAQ AB database that has ceased trading (
> no more OHCL data after say eg. may 2006 ). When  back testing
> sometimes my trading systems do not exit these stocks leaving them in
> an OPEN POSITION till the end of period.
> The problem is that this stock would have been exited in MAY06 and
> another trade taken with available equity .
>
> I can solve this problem by ignoring this stock but it would be better
> if i could exit this STOCK if I could detect it has ceased trading .
>
> Does anyone know how I could code this or ideas ?
>
> Regards
> Moztrader
>

You can add an exit condition that if the last bar is not current date then
exit
eg

sell = sell or ( barindex()==(barcount-1) and lastvalue(datenum())<1061003
);

this is just an exampe, you will need to manage the actual value of the last
date.

--
Cheers
Graham
AB-Write >< Professional AFL Writing Service
Yes, I write AFL code to your requirements
http://e-wire.net.au/~eb_kavan/ab_write.htm

On 04/10/06, intermilan04 <intermilan04@...> wrote:
>
> There may be a way to do what you want to do, but I just delete the
> symbol altogether.  Your system should not stop working just because
> one symbol was removed.
>
> intermilan04
>
> --- In amibroker-ts@yahoogroups.com, "moztrader" <das300@...> wrote:
> >
> > HI,
> > I have stock data in my NASDAQ AB database that has ceased trading (
> > no more OHCL data after say eg. may 2006 ). When  back testing
> > sometimes my trading systems do not exit these stocks leaving them in
> > an OPEN POSITION till the end of period.
> > The problem is that this stock would have been exited in MAY06 and
> > another trade taken with available equity .
> >
> > I can solve this problem by ignoring this stock but it would be better
> > if i could exit this STOCK if I could detect it has ceased trading .
> >
> > Does anyone know how I could code this or ideas ?
> >
> > Regards
> > Moztrader
> >
>
>
>
>
>
>
>
>
>
> Yahoo! Groups Links
>
>
>
>
>
>
>
>
>
>
>


[Non-text portions of this message have been removed]

Moztrader,



I use CSI Data which has extinct stocks, a  "must have" feature IMO. I
use the following code to exit / not enter positions if the stock is
about to be extinct. Unfortunately this does not work real-time because
AB evaluates one stock at a time and the last bar -- even if it's two
years old -- is considered to be the "current date". This also means
that ALL STOCKS appear to be going extinct on their last bar even if
that bar is TODAY! (Yes, I tried comparing dates to NOW(), but that
didn't work either and I don't recall why at this point.)



Hence, I keep two databases (a trick learned from b519b who also uses
CSI Data). The code recognizes which database I am in and uses the
"extinct" detection for my backtesting data and does not use it for R/T
data.



If anyone (like Tomasz) can tell me a way around this, I would be
grateful!



There is also code below to locate the actual price of the stock, which
is another "must have" feature of back-testing stocks.





/****************** BASIC NEEDS of CSI DATA ******************/



dbName = GetDatabaseName();

//>>>>>>>>>>>>>>>>>>>>>>>>>> 1st future reference - looks ahead to find
extinct stocks just to terminate those trades

ThisIsLastBar = BarIndex() == LastValue( BarIndex() ); //Finds the last
bar of data for current Symbol

//>>>>>>>>>>>>>>>>>>>>>>>>>> 2nd future reference - Only use for Explore
since it looks to tomorrow to find the exit price

exitPrice = IIf(ThisIsLastBar, C, Ref(O,1));

if (StrLeft(dbName,3) == "CSI") Price = OI/100; else Price = C; //Always
use actual price on CSI Databases. Other db use the adjusted Close.

if (dbName == "CSI_Research") //Always filter extinct stocks on
CSI_Research data

{

     GoingInactiveIn2Bars = (Ref(ThisIsLastBar, 2) AND Ref(DateNum(), 2)
!= Now(3)) * 16; //Looks ahead 2 bars to see if no more data. Future
look will MATCH current DateNum if last bar is actually the current day.

}

else

{

     GoingInactiveIn2Bars = False;

}





Active = BarIndex() > 50 AND NOT GoingInactiveIn2Bars; //Wait X days for
the first signal AND don't enter if stock is about to go away.



Note: You can use the last variable "Active" in your Buy/Sell statements
to avoid trading stocks that are not "Active".

--

Terry

-----Original Message-----
From: amibroker-ts@yahoogroups.com [mailto:amibroker-ts@yahoogroups.com]
On Behalf Of intermilan04
Sent: Tuesday, October 03, 2006 11:19
To: amibroker-ts@yahoogroups.com
Subject: [amibroker-ts] Re: testing STOCK data that has ceased to trade



There may be a way to do what you want to do, but I just delete the
symbol altogether. Your system should not stop working just because
one symbol was removed.

intermilan04

--- In amibroker-ts@ <mailto:amibroker-ts%40yahoogroups.com>
yahoogroups.com, "moztrader" <das300@...> wrote:
>
> HI,
> I have stock data in my NASDAQ AB database that has ceased trading (
> no more OHCL data after say eg. may 2006 ). When back testing
> sometimes my trading systems do not exit these stocks leaving them in
> an OPEN POSITION till the end of period.
> The problem is that this stock would have been exited in MAY06 and
> another trade taken with available equity .
>
> I can solve this problem by ignoring this stock but it would be better
> if i could exit this STOCK if I could detect it has ceased trading .
>
> Does anyone know how I could code this or ideas ?
>
> Regards
> Moztrader
>





[Non-text portions of this message have been removed]

Hello,

>Unfortunately this does not work real-time because
> AB evaluates one stock at a time and the last bar -- even if it's two
> years old -- is considered to be the "current date".

Hmm... I don't quite understand. If given symbol is two years old then
LastBar( DateNum() ) will be that of TWO YEARS ago, and definitely not
"current".

(unless you use "Pad and align" feature - then "LAST BAR" would be the one of
the reference
symbol, not the "current symbol" one. )

So to detect inactive symbol for say 10 days you can use:

lastday = LastValue( DayOfYear() + 365 * Year() );
nowday = Now( 10 ) + 365 * Now( 8 );

if( nowday - lastday > 10 )
{
   // this symbol is not trading for at least 10 calendar days
}

Best regards,
Tomasz Janeczko
amibroker.com
----- Original Message -----
From: "Terry" <MagicTH@...>
To: <amibroker-ts@yahoogroups.com>
Sent: Wednesday, October 04, 2006 12:19 AM
Subject: RE: [amibroker-ts] Re: testing STOCK data that has ceased to trade


> Moztrader,
>
>
>
> I use CSI Data which has extinct stocks, a  "must have" feature IMO. I
> use the following code to exit / not enter positions if the stock is
> about to be extinct. Unfortunately this does not work real-time because
> AB evaluates one stock at a time and the last bar -- even if it's two
> years old -- is considered to be the "current date". This also means
> that ALL STOCKS appear to be going extinct on their last bar even if
> that bar is TODAY! (Yes, I tried comparing dates to NOW(), but that
> didn't work either and I don't recall why at this point.)
>
>
>
> Hence, I keep two databases (a trick learned from b519b who also uses
> CSI Data). The code recognizes which database I am in and uses the
> "extinct" detection for my backtesting data and does not use it for R/T
> data.
>
>
>
> If anyone (like Tomasz) can tell me a way around this, I would be
> grateful!
>
>
>
> There is also code below to locate the actual price of the stock, which
> is another "must have" feature of back-testing stocks.
>
>
>
>
>
> /****************** BASIC NEEDS of CSI DATA ******************/
>
>
>
> dbName = GetDatabaseName();
>
> //>>>>>>>>>>>>>>>>>>>>>>>>>> 1st future reference - looks ahead to find
> extinct stocks just to terminate those trades
>
> ThisIsLastBar = BarIndex() == LastValue( BarIndex() ); //Finds the last
> bar of data for current Symbol
>
> //>>>>>>>>>>>>>>>>>>>>>>>>>> 2nd future reference - Only use for Explore
> since it looks to tomorrow to find the exit price
>
> exitPrice = IIf(ThisIsLastBar, C, Ref(O,1));
>
> if (StrLeft(dbName,3) == "CSI") Price = OI/100; else Price = C; //Always
> use actual price on CSI Databases. Other db use the adjusted Close.
>
> if (dbName == "CSI_Research") //Always filter extinct stocks on
> CSI_Research data
>
> {
>
>    GoingInactiveIn2Bars = (Ref(ThisIsLastBar, 2) AND Ref(DateNum(), 2)
> != Now(3)) * 16; //Looks ahead 2 bars to see if no more data. Future
> look will MATCH current DateNum if last bar is actually the current day.
>
> }
>
> else
>
> {
>
>    GoingInactiveIn2Bars = False;
>
> }
>
>
>
>
>
> Active = BarIndex() > 50 AND NOT GoingInactiveIn2Bars; //Wait X days for
> the first signal AND don't enter if stock is about to go away.
>
>
>
> Note: You can use the last variable "Active" in your Buy/Sell statements
> to avoid trading stocks that are not "Active".
>
> --
>
> Terry
>
> -----Original Message-----
> From: amibroker-ts@yahoogroups.com [mailto:amibroker-ts@yahoogroups.com]
> On Behalf Of intermilan04
> Sent: Tuesday, October 03, 2006 11:19
> To: amibroker-ts@yahoogroups.com
> Subject: [amibroker-ts] Re: testing STOCK data that has ceased to trade
>
>
>
> There may be a way to do what you want to do, but I just delete the
> symbol altogether. Your system should not stop working just because
> one symbol was removed.
>
> intermilan04
>
> --- In amibroker-ts@ <mailto:amibroker-ts%40yahoogroups.com>
> yahoogroups.com, "moztrader" <das300@...> wrote:
>>
>> HI,
>> I have stock data in my NASDAQ AB database that has ceased trading (
>> no more OHCL data after say eg. may 2006 ). When back testing
>> sometimes my trading systems do not exit these stocks leaving them in
>> an OPEN POSITION till the end of period.
>> The problem is that this stock would have been exited in MAY06 and
>> another trade taken with available equity .
>>
>> I can solve this problem by ignoring this stock but it would be better
>> if i could exit this STOCK if I could detect it has ceased trading .
>>
>> Does anyone know how I could code this or ideas ?
>>
>> Regards
>> Moztrader
>>
>
>
>
>
>
> [Non-text portions of this message have been removed]
>
>
>
>
> Yahoo! Groups Links
>
>
>
>
>
>
>
>
>
>
>