|
PureBytes Links
Trading Reference Links
|
Bob:
Speaking in general terms: no 2 algorithms will give the same answer. Back
in the '60s or early '70s I saw a comparison of algorithms and computers.
The author took 3 equations in 3 unknowns (x, y , and z) and solved them
using different algorithms on the same computer and the same algorithms on
different computers. I dont remember the exact numbers, but some of the
results were off by as much as 50%, some were quite close. The author used
x = y = z = 1
Since the scans and the chart would/should be using the same algorithm, they
should be identical. Since they are not, there is a defect in the
Quotes-Plus program.
The macd uses exponential moving averages. The value of the moving averages
will depend on how much data is used. If you are using more than about 5
times the long moving average, about 200 days, the differences should be
small enough to ignore.
Lionel Issen
-----Original Message-----
From: Bob Jagow <bjagow@xxxxxxx>
To: metastock@xxxxxxxxxxxxx <metastock@xxxxxxxxxxxxx>
Date: Tuesday, November 24, 1998 1:57 AM
Subject: Re: QP2 / metastock calculation differences
>Should be identical, Ed, once you base both of them on the same # of days.
>QP2 loads all available data.
>
>
>At 10:47 PM 11/23/98 -0500, you wrote:
>>To all,
>>I am only posting to this board because so many here are using QP2 for
their
>>data supply. I have noticed a calculation problem which I would like
>>someone to verify before I contact QP2. This has to do with the MACD
>>calculations. For example:
>>
>>Stock Symbol: ALLE
>>Metastock: MACD (0) = -0.1991, MACDsignal(0) = -0.2496
>> MACD (-1) = -0.2581, MACDsignal(-1) = -0.2622
>>
>>QPChart: MACD (0) = -0.2084, MACDsignal(0) = -0.2624
>> MACD (-1) = -0.2712, MACDsignal(-1) = -0.2759
>>
>>QPScan: MACD (0) = -0.1946, MACDsignal(0) = -0.2426
>> MACD (-1) = -0.2563, MACDsignal(-1) = -0.2546
>>
>> My reason for the concern is the crossover signal date. In this case the
>>signal is triggered in the QPScan as of today when the QPchart and
Metastock
>>has the signal triggered at least one or more days before. Can anyone
>>verify these numbers and suggest if there is a problem or not? I don't
want
>>to alarm QP if not needed.
>>
>>Secondly, why are all the numbers not exact? The calculation is not that
>>complicated and all software is using the same data.
>>
>>thanks in advance,
>>
>>
>>
>
>Bob Jagow
>bjagow@xxxxxxx
>
|