|
PureBytes Links
Trading Reference Links
|
Steve,
I use Fibonacci numbers in a lot of my system tests. I don't
know why, but they seem to work better.
Jim
-----Original Message-----
From: Steve Karnish <kernish@xxxxxxxxxxxx>
To: metastock@xxxxxxxxxxxxx <metastock@xxxxxxxxxxxxx>
Date: Friday, July 17, 1998 8:20 PM
Subject: Re: Tema
>Jim,
>I noticed in your advice to Bill that you used the numbers: 21,
>34, 55. What part does Fibonacci numbers play in your overall
>anlaysis? Is anyone out there going beyond the arcs,
>retracements, and period tools that MS provides?
>Steve Karnish
>CCT
>
>----------
>> From: Jim Greening <JimGinVA@xxxxxxxxxxxxx>
>> To: metastock@xxxxxxxxxxxxx
>> Subject: Re: Tema
>> Date: Friday, July 17, 1998 1:04 PM
>>
>> Bill,
>> That's right, what you have is a system test formula.
>The
>> explorer can't optimize, where the system tester can. To use
>the
>> formula in the Explorer, you have to replace opt2. Start out
>using
>> 21, then you can try the other values of 34 or 55 to see which
>you
>> like best. In fact you can experiment with any number you
>want.
>>
>> Jim
>> -----Original Message-----
>> From: Bill G Bolumen <BOLUMENB@xxxxxxxxxxx>
>> To: metastock@xxxxxxxxxxxxx <metastock@xxxxxxxxxxxxx>
>> Date: Thursday, July 16, 1998 6:24 PM
>> Subject: Re: Tema
>>
>>
>> >Jim,
>> >I am not good at all with formulas and I was trying to copy
>your
>> Short
>> >Formula into explorer and it says that "optimization
>variables are
>> not
>> >allowed in this type of formula"
>> >What else do I need to other than just inputting the short
>formula
>> into the
>> >explorer?
>> >Thanks,
>> >Bill
>> >
>> >----------
>> >> From: Jim Greening <JimGinVA@xxxxxxxxxxxxx>
>> >> To: metastock@xxxxxxxxxxxxx
>> >> Subject: Re: Tema
>> >> Date: Wednesday, July 15, 1998 6:54 PM
>> >>
>> >> CMA,
>> >> I'll copy an old post below.
>> >>
>> >> Jim
>> >> --------------------------------------
>> >> The Rsquared, S/C,MFI test is based on linear
>regression.
>> >> RSquared is a function that measures how strong a data
>array of
>> given
>> >> length is trending. A return of one correlates to a very
>strong
>> trend
>> >> in either direction. A return of zero correlates to no
>trend. I'm
>> >> using it to tell me when there is no trend over the last 21
>days,
>> or
>> >> that rsquared(C,21) is less than 0.15. Now that we have no
>trend,
>> we
>> >> would expect one to start in the next 5 to 13 days. If one
>does
>> start
>> >> we want to determine the direction and get in early.
>Therefore
>> >> Rsquared is used as a trigger, using it as an ALERT
>function. With
>> >> the ALERT given, if all the rest of the conditions are met
>any time
>> >> within a 13 day period, the buy signal will be generated.
>I'm then
>> >> using S/C and MFI to determine trend start and direction.
>> >> S/C is nothing more than the normalized 34 day
>LinRegSlope of
>> the
>> >> close. I normalize it by dividing it by the close so
>charts are
>> more
>> >> directly comparable. I also multiply by 10000 just to get
>numbers
>> >> between 0 and a few hundred. Finally I apply Tema
>smoothing to get
>> a
>> >> smoother plot. The theory is that if S/C is rising then a
>uptrend
>> is
>> >> in progress. If S/C is falling a down trend is in
>progress. I
>> check
>> >> this by making sure that tema(S/C) is larger than an
>optimized
>> value
>> >> (opt1) and is rising. I use the function HHV(X,5) =
>HHV(X,13) to
>> >> check. This is just a quick way to determine if the most
>reason
>> >> values of X have been rising. If the condition is met,
>then S/C is
>> >> rising. Similarly if it met the condition LLV(X,5) =
>LLV(C,13) it
>> >> would be falling.
>> >> I double check the result by also making sure that MFI
>is also
>> >> rising. When all conditions are met within 13 days of the
>alert,
>> we
>> >> get a buy signal. The converse is true for a sell signal.
>Since
>> the
>> >> market is biased upwards, I add two additional checks for
>the
>> short -
>> >> MFI less then 0 and the 55 day variable moving average
>falling. I
>> >> close the positions when the trend has stopped. For the
>long
>> position
>> >> I check that the Tema(MFI) is below 0, the Tema(S/C) is
>below opt 1
>> >> and a 55 day exp moving average is falling. Short
>positions tend
>> to
>> >> move faster, so I only look the 144 day moving aver to
>begin rising
>> to
>> >> close that position. The complete test follows:
>> >>
>> >> ENTER LONG:
>> >> Alert(RSquared(C,21) < 0.15,13) AND
>> >> Tema(10000*LinRegSlope(C,34)/C,34) > opt1 AND
>> >> HHV(Tema(10000*LinRegSlope(C,34)/C,34),5) =
>> >> HHV(Tema(10000*LinRegSlope(C,34)/C,34),13) AND
>> >> HHV(Tema(MFI(55),55),5) = HHV(Tema(MFI(55),55),13)
>> >>
>> >> CLOSE LONG:
>> >>
>> >> Tema(MFI(55),55) - 50 < 0 AND
>> >> Tema(10000*LinRegSlope(C,34)/C,34) < opt1 AND
>> >> LLV(Tema(10000*LinRegSlope(C,34)/C,34),5) =
>> >> LLV(Tema(10000*LinRegSlope(C,34)/C,34),13) AND
>> >> LLV(Tema(MFI(55),55),5) = LLV(Tema(MFI(55),55),13)
>> >>
>> >> ENTER SHORT
>> >>
>> >> Alert(RSquared(C,21) < 0.15,13) AND
>> >> Tema(10000*LinRegSlope(C,34)/C,34) < opt2 AND
>> >> LLV(Tema(10000*LinRegSlope(C,34)/C,34),5) =
>> >> LLV(Tema(10000*LinRegSlope(C,34)/C,34),13) AND
>> >> LLV(Tema(MFI(55),55),5) = LLV(Tema(MFI(55),55),13) AND
>> >> Tema(MFI(55),55) - 50 < 0 AND
>> >> LLV(Mov(C,55,VAR),5) = LLV(Mov(C,55,VAR),13)
>> >>
>> >> CLOSE SHORT
>> >>
>> >> HHV(Mov(C,55,VAR),5) = HHV(Mov(C,55,VAR),13) AND
>> >> Tema(MFI(55),55) - 50 > 0
>> >>
>> >> OPTIMIZATION:
>> >>
>> >> OPT1: Min=-34 Max=-8 Step=13
>> >> OPT2: Min=-55 Max=-21 Step=34
>> >>
>> >> -----Original Message-----
>> >> From: amc amc <cma6@xxxxxxxxx>
>> >> To: metastock@xxxxxxxxxxxxx <metastock@xxxxxxxxxxxxx>
>> >> Date: Wednesday, July 15, 1998 11:09 AM
>> >> Subject: Tema
>> >>
>> >>
>> >> >
>> >> >To Jim Greening:Don't know if you have covered this, but
>can you
>> >> >describe Tema S/C and Tema MFI (money flow index?),
>howconstructed
>> >> >and how interpreted. Thanks, CMA
>> >> >
>> >> >
>> >> >
>> >> >_________________________________________________________
>> >> >DO YOU YAHOO!?
>> >> >Get your free @yahoo.com address at http://mail.yahoo.com
>> >> >
>> >>
>> >>
>>
>>
|