|
PureBytes Links
Trading Reference Links
|
Kevin
How many periods are you using for your exploration? If you are using
"Minimum" then you should expect exactly the problem you have.
RSI uses Wilders Smoothing internally, which is a form of exponential moving
average. You need to force your exploration to use about 75-100 periods to
get accurate results from RSI(14).
There are over 20 other MS functions that will exhibit the same problem so
beware.
Roy
----- Original Message -----
From: "Kevin Howard" <kjhoward@xxxxxxxxxxxxxx>
To: <Metastockusers@xxxxxxxxxxxxxxx>
Sent: Friday, April 25, 2003 11:19 PM
Subject: [Metastockusers] Exploration RSI(14) Problems
> Hi Manish and Preston,
>
> Thank you both for offering help with my RSI
> (14) Exploration problems. I have spent the best part of today trying to
> move forward but I am still identifying false positives. That is I am
> still identifying shares that should have crossed above the RSI (14) 30
> level within the last 5 trading days, but in fact have not been below
> RSI (14) 30 during this period. I only have limited MS programming
> skills but I have happily used MS for the past 5 years and I have used
> my own explorations successfully for much of this time. However I would
> say I am a little shaken with the problem I am currently experiencing.
>
> Taking each of your posts in turn, I pasted your
> Exploration code Manish and it ran smoothly but did not eliminate my
> "false positive" problem. I am afraid I was not able to work out the
> logic behind what looks to be a nice piece of code, so I could not take
> it any further. Could you briefly how you exploration works?
>
> Preston I picked up on your suggestion re
> validation but this has only deepened my concern. I had previously
> looked upon validation as being required when there was a possibility
> that there may be a logic conflict within an Exploration code. I would
> have thought that Alert (Cross (RSI (14), 30), 5) was fairly
> unambiguous, (in fact the 'RSI cross 70' is given as an "Alert" example
> on Page 220 of my User Manual), but apparently not. Nevertheless I have
> now used the following as an attempt to validate:
>
> A: =Alert (Cross (RSI (14), 30), 5);
> B: =If (BarsSince (A) <=5, 1, 0); B=1
>
> But this still gave me false positives, i.e. shares that had not been
> below 30 within the past 5 trading days. So I ran another exploration
> this time using the following filter:
>
> BarsSince (RSI (14) <30) <=5 AND
> BarsSince (RSI (14)>30) <=5
>
> I also used my columns A to F to record the RSI (14) values for each of
> the past 5 trading days using the 'ref' function. To my astonishment and
> despair I still captured shares whose RSI (14) value had not been below
> 30 during the past 5 trading days. So is my logic wrong with these
> attempts to validate or have I somehow corrupted my MS software or data?
>
> I should add that I definitely take your point about trading whipsaws.
> The reason why I am so pedantic about wanting to capture shares only
> where either RSI (14) has crossed 30 or MACD () signals 'Buy' and
> desirably to have both occur within the last few trading days is that I
> then want to overlay a composite template over such selected shares,
> interpret the resultant graph and then act on this interpretation. The
> RSI and MACD signals do not form part of the template but are highly
> correlated in situations where the template signals a buy or sell
> action. On a small sample of shares (5) in real life, the template has
> given exceptionally good and unambiguous results. However as I cannot
> employ all the parameters used in the template as an exploration or
> System Tester code, I am left with having to manually test it on a
> larger number of potential candidate shares before I get too carried
> away with its usefulness as a trading tool.
>
> Two final points: I too thought that Ref -1 gave yesterday's value and
> Ref -2 gave the trading day before. But now I strongly suspect that it
> gives the value of the last and second last day that a particular share
> or derivative traded. For example in the first validation attempt used
> above, I used two of the A-F column to record the DayOfMonth () and ref
> (dayofmonth (),-1) values. I fully expected to see 24 and 23 returned
> respectively (25 is a public holiday in Australia). The majority
> reflected these values but a substantial minority did not, but did
> appear to reflect the last and second last days they traded. So I would
> welcome clarification on the reference function. And finally I still
> would very much like to filter out derivatives and retain straight
> stock. So if there is anyway of filtering on the Ticker Symbol character
> size within Metastock, I would greatly appreciate being told about it.
>
> I really am finding this RSI (14) a difficult problem to solve and yet I
> know it must be that I am doing something or some things wrongly. So any
> further insights from either of you or any others on the list would be
> really well received here.
>
> Thanks again, Kevin Howard
>
>
>
>
>
>
>
------------------------ Yahoo! Groups Sponsor ---------------------~-->
Get A Free Psychic Reading! Your Online Answer To Life's Important Questions.
http://us.click.yahoo.com/O10svD/Me7FAA/AG3JAA/zMEolB/TM
---------------------------------------------------------------------~->
To unsubscribe from this group, send an email to:
Metastockusers-unsubscribe@xxxxxxxxxxx
Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
|