PureBytes Links
Trading Reference Links
|
Hi Tomasz,
Why does AB treat the no trades on a stock condition as an exception when
calculating B&H?
If for example a trading system fails to generate any long trades during the
test period when in fact the stock has risen considerably during that test
period then surely the performance measurement of the system should reflect
this (i.e.system has lost out compared to Buy & Hold strategy).
Regards
John
----- Original Message -----
From: "Tomasz Janeczko" <amibroker@xxxx>
To: <amibroker@xxxxxxxxxxxxxxx>
Sent: Wednesday, June 06, 2001 7:47 AM
Subject: Re: [amibroker] Re: buy and hold results change
> Hi.
>
> I am sorry for not answering this question for some days but
> I was really busy working on the new version.
> Now to the buy/hold performance calculation -
> your guesses are somewhat wrong:
> it does take first and the last day of the period
> specified (either by dates or number of quotes).
> It does not care about when the first signal in your
> formula appears.
> There is one exception however - if your formula
> does NOT generate ANY trade within specified
> period - buy and hold result for that stock is
> not counted at all. It means when you back test 10 stocks
> and your formula generates trades for 9 stocks but
> no trade for 1 stock - overall buy and hold results will
> be calculated also for that 9 stocks.
>
> Note: from version 3.6 buy and hold results are calculated
> using buyprice and sellprice arrays so if your formula
> assigns the values dynamically to this arrays or you
> change the settings of fields used for buy/sell in analysis
> settings - this will affect buy/hold results.
>
> So general formula for buy /hold is:
>
> firstbar = the first quote choosen by you in "Range" settings
> lastbar = the last quote choosen by you in "Range" settings
>
> if( anytrades )
> {
> BuyHoldProfit = SellPrice[ lastbar ] - BuyPrice[ firstbar ]
> }
>
>
>
> As a proof take the following three formulas:
> 1.
>
> buy = cum(1)== 30;
> sell = cum(1 ) == 120;
>
> 2.
> buy = cum(1)== 50;
> sell = cum(1 ) == 110;
>
> 3.
> buy = cum(1)== 60;
> sell = cum(1 ) == 90;
>
> If only you have enough quotes > 120 and you select the range
> that bar 30 and 120 is within the range (all quotations for example will
do)
> the buy/hold result remains the same regardless of moving entry/exit dates
> of the formula
>
>
> Best regards,
> Tomasz Janeczko
> ===============
> AmiBroker - the comprehensive share manager.
> http://www.amibroker.com
>
> ----- Original Message -----
> From: "Tom McDaniel" <tmtempe@xxxx>
> To: <amibroker@xxxxxxxxxxxxxxx>
> Sent: Wednesday, June 06, 2001 7:13 AM
> Subject: Re: [amibroker] Re: buy and hold results change
>
>
> > David-
> >
> > There may not be a perfect solution. If it takes 6 months for the first
buy
> > signal to be generated by the system under test and the buy/hold has
already
> > appreciated significantly I think that seems important to know in doing
a fair
> > comparison to buy/hold.
> >
> > Further, I think the original question related to comparing two systems
that had
> > different times for the first signal, say three months and six months.
By your
> > method they each have a different buy/hold value that they are measured
against
> > -- even though, I believe, they are really measured against the exact
same data
> > set.
> >
> > For rigorous analysts, one should not start an analysis until the date
when all
> > indicators or devices (e.g. moving averages, etc., if they are used)
have "grown
> > in" to correct values.
> >
> > In summary, I would like for the buy to take place on the date that I
set as the
> > start date and the end date to be the final price for calculation
purposes. For
> > each analysis over the same time period the buy/hold would provide the
same
> > result regardless of when the first system trade takes place.
> >
> > Regards,
> > -Tom McDaniel
> >
> > David Holzgrefe wrote:
> >
> > > Hi John R and others
> > >
> > > I have been thinking about this
> > > And have come to the conclusion the way IHMO AB calculates the
buy/hold is
> > > correct .
> > >
> > > If on day 1 your buy and hold to end date this is the result
> > > now if you wish to see how your indicator performs compared to
buy/hold
> > >
> > > You could run the scan 1 st then look for the 1st date that the afl
signals
> > > a buy then adjust your testing period to suit .
> > >
> > > That way you would get a better picture of how it performed over the
period
> > > compared to buy/hold .
> > >
> > > David
> > >
> > > ----- Original Message -----
> > > From: "David Holzgrefe" <dtholz@xxxx>
> > > To: <amibroker@xxxxxxxxxxxxxxx>
> > > Sent: Wednesday, June 06, 2001 10:04 AM
> > > Subject: Re: [amibroker] Re: buy and hold results change
> > >
> > > > John R
> > > > I Think I follow you suggestion/ inquiry
> > > >
> > > > the Buy/hold starts the calculation from the first period in the
range
> > > > condition ,
> > > > When if fact the AFL will may not start calculations for say xn
periods
> > > > after the date.
> > > >
> > > > Giving a unequal comparison ?
> > > >
> > > > an example would be ..
> > > > we set our search for a cross over of a mov but because the stock in
a
> > > > steady up trend the mov afl is not triggered for several months when
a
> > > small
> > > > pull back occurs ..trigging the afl but the buy/hold has started
counting
> > > > the returns from day 1 of the range period ..
> > > >
> > > > A valid point John I haven't look to confirm if this is the way AB
does
> > > the
> > > > calculation.
> > > > This would explain why its hard to best the buy/hold .
> > > >
> > > > included is a image of an example search
> > > > some unexpected results maybe the sell = sell OR lastbar; doesn't
get
> > > > counted in the calculations for the afl ?
> > > >
> > > > Please excuse me if I'm barking up the wrong tree ( got it wrong)
> > > > but I thought I would post a detailed answer so as newer ppl could
get a
> > > > hold on the question .
> > > >
> > > > range was set to 12 mth 06.06.2000 to06.06.2001
> > > >
> > > > TJ will know
> > > >
> > > >
> > > > Regards David
> > > >
> > > >
> > > > ----- Original Message -----
> > > > From: "John R" <jrdrp@xxxx>
> > > > To: <amibroker@xxxxxxxxxxxxxxx>
> > > > Sent: Wednesday, June 06, 2001 8:40 AM
> > > > Subject: Re: [amibroker] Re: buy and hold results change
> > > >
> > > >
> > > > > IMO Buy & Hold calculations should be based on prices at start and
end
> > > > dates
> > > > > of date range under test. Using date of first signal would not
give
> > > > results
> > > > > which could be used to fairly compare different systems fairly.
Consider
> > > > for
> > > > > example the case of a system which does not generate its' first
buy
> > > signal
> > > > > until say 6 months into the period under test, during which stock
has
> > > > > already risen say 20%.
> > > > >
> > > > > John
> > > > > ----- Original Message -----
> > > > > From: <cliffelion@xxxx>
> > > > > To: <amibroker@xxxxxxxxxxxxxxx>
> > > > > Sent: Tuesday, June 05, 2001 3:12 PM
> > > > > Subject: [amibroker] Re: buy and hold results change
> > > > >
> > > > >
> > > > > > Hi David,
> > > > > >
> > > > > > I am using the same timeframe, stocks and analysis settings.
When I
> > > > > > change systems the analysis results change as expected - but so
do
> > > > > > the buy and hold results. I want to use the buy and hold as the
> > > > > > baseline to compare the different systems so I'm not sure how to
> > > > > > compare the systems when the buy and hold results change.
> > > > > >
> > > > > > AMI may be using the buy point of the system to start the buy
and
> > > > > > hold calculation - which means that the system is influencing
the buy
> > > > > > and hold.
> > > > > >
> > > > > > This may be OK - but I would be interested to know if there are
other
> > > > > > methods for comparing systems.
> > > > > >
> > > > > > Regards,
> > > > > >
> > > > > > Cliff
> > > > > >
> > > > > >
> > > > > >
> > > > > > --- In amibroker@xxxx, "David Holzgrefe" <dtholz@xxxx> wrote:
> > > > > > > I'm not sure if I follow what you mean Cliff
> > > > > > > But if you change the system triggers? then I would expect
that the
> > > > > > buy sell
> > > > > > > condition would change.
> > > > > > >
> > > > > > >
> > > > > > > Regards David
> > > > > > >
> > > > > > > ----- Original Message -----
> > > > > > > From: <cliffelion@xxxx>
> > > > > > > To: <amibroker@xxxx>
> > > > > > > Sent: Tuesday, June 05, 2001 4:40 PM
> > > > > > > Subject: [amibroker] buy and hold results change
> > > > > > >
> > > > > > >
> > > > > > > > When I perform back tests over the same time period with the
same
> > > > > > > > stocks, I find that the buy and hold results change when I
use
> > > > > > > > different systems.
> > > > > > > >
> > > > > > > > I assume that this is because the buy and hold starts
calculating
> > > > > > > > when the system triggers a buy.
> > > > > > > >
> > > > > > > > Does anyone know if this is true???
> > > > > > > >
> > > > > > > > Thanks
> > > > > > > >
> > > > > > > > Cliff Elion
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > > Your use of Yahoo! Groups is subject to
> > > > > > http://docs.yahoo.com/info/terms/
> > > > > > > >
> > > > > > > >
> > > > > >
> > > > > >
> > > > > >
> > > > > >
> > > > > > Your use of Yahoo! Groups is subject to
> > > > http://docs.yahoo.com/info/terms/
> > > > > >
> > > > > >
> > > > >
> > > > >
> > > > >
> > > > >
> > > > > Your use of Yahoo! Groups is subject to
> > > http://docs.yahoo.com/info/terms/
> > > > >
> > > > >
> > > >
> > > >
> > > >
> > > > Your use of Yahoo! Groups is subject to
http://docs.yahoo.com/info/terms/
> > > >
> > > >
> > >
> > >
> > >
> > > Your use of Yahoo! Groups is subject to
http://docs.yahoo.com/info/terms/
> >
> >
> >
> >
> >
> > Your use of Yahoo! Groups is subject to
http://docs.yahoo.com/info/terms/
> >
> >
> >
>
>
>
>
>
> Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
>
>
|