[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [amibroker] Re: Winning Systems



PureBytes Links

Trading Reference Links

Dimitris-

Interesting you should mention this. I already did just what you are
suggesting and found a HUGE difference in the backtest performance (all other
conditions identical). Although I did not have time to pursue and understand
it, the difference seemed to be much larger than could be explained by just a
one day delay in entering a position.

Any thoughts?

-Tom

DIMITRIS TSOKAKIS wrote:

> Tom,
> There is something that I do not understand.
> You do not know ref(close,1) and you accept for the
> last day the convention ref(close,1) to be equal to
> lastvalue(close), which is not true of course.
> Instead of
> mo=Close/Ref(Close,1)*100 - RSI(14);
> you may use
> moo=Ref(Close,-1)/Close*100 - Ref(RSI(14),-1);
> It is real for any day and has no time problem.
> You will notice that moo has 1 day lag compared to
> mo, but mo is useful only for the past and everyday has an
> unrealistic concession for tomorrow, i.e. is a bit useless.
> Does it help ?
> Dimitris Tsokakis
> --- In amibroker@xxxx, Tom McDaniel <tmtempe@xxxx> wrote:
> > Tom-
> >
> > Two versions are listed below: the original posting by Jaco and one
> that
> > I have modified and tweaked (this is a work in progress).
> >
> > I am bothered by the presence of the ref (close, 1) in the
> indicator.
> > As I understand it this represents tomorrow's close (which is
> unknown
> > today!). When doing a scan for today AFL replaces tomorrow's close
> with
> > today's close, which is fine. However, when doing a back test for
> all
> > past dates tomorrow's close is a known quantity.
> >
> > My concerns or questions:
> >
> > * If a buy signal today is based on tomorrow's close being equal
> to
> > today's close will the signal disappear if tomorrow is a down
> day?
> > * Could I get a buy signal yesterday for yesterday that doesn't
> show
> > up until today because today's close is higher than
> yesterday's?
> >
> > I have posted these questions some time back to the group but did
> not
> > get any response.
> >
> > I have found that the CCI with colored bars performs similarly to
> the
> > MRI but the backtested performance is significantly inferior. I
> worry
> > that the difference is due to the ref(close, 1) artifact discussed
> > above.
> >
> > I would appreciate any clarity that anyone can supply.
> >
> > Regards,
> > -Tom McDaniel
> >
> >
> > /* MRI developed by Jaco Jonker */
> >
> > mo=close/ref(close,1)*100 - rsi(14);
> > mo2=ma(mo,100);
> > graph0=mo;
> > graph1=rsi(14);
> >
> >
> >
> > title= "MRI(100):"+name() + writeval( graph0 );
> >
> >
> > /* MRI developed by Jaco Jonker */
> > /* Modified by T. L. McDaniel */
> > /* Initially Optimized to 19 days 7/2/2001*/
> >
> > Maxgraph=5;
> > Time=19;
> > mo=close/ref(close,1)*100 - rsi(Time);
> > mo2=ma(mo,100);
> > graph0Color=5;
> > graph0=mo;
> > Graph1Color=6;
> > graph1=rsi(Time);
> > Graph2style=4;
> > graph2=adx(Time);
> >
> > graph3style=2;
> > graph4style=1;
> > graph3barcolor=iif(rsi(Time)>mo,5,4);
> > graph3=rsi(Time)-mo;
> > graph4=rsi(Time)-mo;
> >
> > Buy=cross(rsi(Time),mo) ;
> > Sell=cross(mo,rsi(Time));
> > Short=cross(mo,rsi(Time));
> > Cover=cross(rsi(Time),mo);
> >
> > title= "DATE REF(Close,1) Close RSI(Time) MO";
> > title=("TLMTS "+ Name() +" " + Date()) +" "+writeval(ref(close,1))
> +" "+writeval(close)+" "+writeval(rsi(Time)) +" "+writeval(mo);
> >
> > T = 20;
> > title = "301-4 " + name () + " " + fullname() + " "
> > + "CCI" + " " + "periods:" + writeval(T);
> > maxgraph = 5;
> > graph2 = CCI (T);
> >
> > Uptrend = 100; Downtrend = -100;
> >
> > graph2barcolor = IIF (graph2 > uptrend, 5,
> > IIF (graph2 < downtrend,4,6));
> >
> >
> > graph4 = graph2;
> >
> > buy = cross (graph2, 0) and (ref (graph2,-1) < 0) and (ref (graph2,-
> 2) < 0)and (ref (graph2,-3) < 0) and (LLV (rsi(9),10) < 30) and cross
> (rsi (9),50);
>
>
>
> Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/