|
PureBytes Links
Trading Reference Links
|
Exactly what I wrote about my poor english:
your third approach was my second proposition:
SC
> Actually, I was thinking of a third approach. Group by
characteristic and then optimize, rather than optimize the universe
and then filter.
>
> Bill
> ----- Original Message -----
> From: Stephane Carrasset
> To: amibroker@xxxx
> Sent: Wednesday, October 03, 2001 9:21 AM
> Subject: [amibroker] Re: Oscillators II (%D)
>
>
> Totally agree with you,
> the stocks's selection is one important steps in a profitable
trading
> system.
> Actually, I don't know what is the best solution to select
tradable
> stocks.
>
> 1/ apply a trading system on a universe and trade only the best
> performers , then make a turnover of the winners .
>
> 2/OR apply a filter on a universe and trade only the stocks with
> this filter
> this filter could be a trend filter ( the choice is large)
> or a statistic filter to know if the distribution of the price is
> random/not random
>
> Stephane carrasset
>
>
>
> > Although I do not use systems for most of my trading, I do use
them
> and agree that testing a universe is more appropriate as it tends
to
> produce a more robust system. In fact, imo, this is one of the
> advantages of AB vs MS. With the universe approach, however,
have
> you looked at the desirability of using subsets? For example,
rather
> than combine stocks into a single universe, combine them into
smaller
> groupings that have a common characteristic such a volatility.
> Intuitively, this would appear to be another step toward
robustness.
> >
> > Bill
> > ----- Original Message -----
> > From: Stephane Carrasset
> > To: amibroker@xxxx
> > Sent: Wednesday, October 03, 2001 8:28 AM
> > Subject: [amibroker] Re: Oscillators II (%D)
> >
> >
> > Dimitri,
> >
> > I am not in favour for an optimization for one stocks,
perhpas on
> a
> > universe of stocks would be better, it is one aim of the
> montecarlo
> > analysis, we'll have it perhpas in the system tester of
> version4.8 of
> > amibroker ^__^
> >
> > But I am thinking it is a good idea of variables OBOS level
for a
> > fixed stochastic with your code below
> > obos level could be + or - 40% above or below the MEANST ( 70
is
> +40%
> > of 50)
> >
> > /*MEAN %D EXPLORATION*/
> > MEANST=cum(stochd())/(cum(1)-18);
> > filter=CUM(1)>100;
> > numcolumns=1;
> > column0=MEANST;
> > column0format=1.0;
> > column0name="MEAN STOCHD";
> >
> >
> > SC
> >
> > > Optimization is neccessary for each stock in order to
> > > find the best X1, Dbuy and Dsell.
> > > For the moment I have not any homework for a universal
> > > value of above parameters Time and experience will show
> > > if it is possible and profitable.
> > > Any co-operative idea appreciated.
> > > Dimitris Tsokakis
> > > --- In amibroker@xxxx, "Stephane Carrasset"
<s.carrasset@xxxx>
> > wrote:
> > > > Dimitry,
> > > >
> > > > Optimization on a trading system means that you have
> backtested
> > on
> > > > only one stock and not an universe of stocks?
> > > >
> > > > without optimization the idea of various OBOS level is
great,
> > > > instead of buy when stoch crosses below an OS level, we
> could
> > buy
> > > > when stoch is < OS level and buy on close delay when H >
ref
> (H,-)
> > > > *1.005
> > > >
> > > > SC
> > > >
> > > > > AVST=MA(StochD(X1),100);
> > > > >
> > > > > Let us try now to buy lower and sell higher.
> > > > > The buy level will be AVST-Dbuy and the sell level will
> > > > > be AVST+Dsell. If you guess now, optimization may give
> > > > > the best values for the 3 parameters.
> > > > > Optimize a current stock for all quotations using the
code
> > > > >
> > > > > /*Buy-Sell moving levels for Slow Stochastic*/
> > > > > Dbuy=Optimize("Dbuy",14,0,20,1);
> > > > > Dsell=Optimize("Dsell",8,0,20,1);
> > > > > X1=Optimize("X1",13,10,20,1);
> > > > > s1=StochD(X1);
> > > > > AVST=MA(s1,100);
> > > > > Buy = Cross( s1,AVST-DBuy);
> > > > > Sell = Cross( AVST+Dsell,s1);
> > > > >
> > > > > Then replace manually the 14, 8, 13 with respective
results
> of
> > > your
> > > > > optimization and scan and back test.
> > > > > Compare the back test result with the traditional
> > > > >
> > > > > buy=cross(stochd(),30);
> > > > > sell=cross(70,stochd());
> > > > >
> > > > > scan and back test.
> > > > >
> > > > > I found interesting net profit augmentation.
> > > > > As for settings, I used buy at low, sell at high with
delay
> 1
> > day
> > > > > (it is obvious that you get the signal today and you
act
> > tomorrow
> > > > > and it is always important for back testing. Delay 0 is
> > > meaningless,
> > > > > the signal is first and then the action may be done.)
> > > > > After the calculation of above parameters, you may see
in
> your
> > > > > Indicator builder the new curves, pasting the formula:
> > > > >
> > > > > /*Slow Stochastic moving buy-sell levels*/
> > > > >
> > > > > Dbuy=14;
> > > > > Dsell=8;
> > > > > X1=13;
> > > > > MaxGraph=12;
> > > > > ST3=StochK(X1);
> > > > > ST33=StochD(X1);
> > > > > Graph0=ST3;
> > > > > Graph1=ST33;
> > > > > AVST=MA(StochD(X1),100);
> > > > > Graph9=AVST-Dbuy;Graph10=AVST+Dsell;
> > > > > Graph8=avst;Graph8Style=8;Graph8BarColor=2;
> > > > > Graph9Style=Graph10Style=8;
> > > > > Graph9BarColor=Graph10BarColor=1;
> > > > >
> > > > > and replace 14, 8, 13 with the results of optimization.
> > > > > Real examples will follow.
> > > > > Dimitris Tsokakis
> >
> >
> >
> >
> >
> > 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/
|