shelupinin
120 posts
msg #37950
- Ignore shelupinin |
9/14/2005 7:16:42 AM
below is just another modification of previous filters, try to see performance for last 1-10 days period, almost all days are GREEN !
|
shelupinin
120 posts
msg #37951
- Ignore shelupinin |
9/14/2005 7:23:12 AM
for the previous filter, up to 15 days holding period most days are GREEEN
|
yepher
359 posts
msg #37954
- Ignore yepher |
9/14/2005 11:33:30 AM
shelupinin,
Much of those reports are made manually right now. I am using MetaStock, Stockfetcher, and some custom homegrown tools.
As for the "mostly GREEN" part of your previous message. You have to be *really careful* with SF’s "performance" tool. It shows results as if you were to buy the stock on the trigger day. This in my opinion is misleading for an EOD tool like SF. SF is a really GREAT tool which has many strengths (price, speed, simplicity, etc). But they are not perfect at everything and I believe back testing is not yet good. The good thing is they are putting resources on the problem and are getting better.
My Modus operandi is to work with SF to rapidly prototype filters. Convert them to MetaStock for back testing and refinement. Then us SF for day to day stock selection once the filter has been characterized.
To help the situation you can also use a technique originally formed by Cegis and used/refined by RumpledOne. You can find it documented here http://yepher.com/~yepher/stockfetcher/command.html under the “Back Testing” section. This is one of two items in that document that are user provided. (the other is RumpledOnes logic system).
Incidentally, I would like to add a few more because I see some indicators on this forum that are suffering from copy and past logic errors. Candle Shadow is one of them. Here is an example of the right way to do it: http://www.stockfetcher.com/stockdb/fetcher?p=forum&sub=view&fid=1010&tid=36353.
I hope that helps,
-- Yepher
|
nikoschopen
2,824 posts
msg #37960
- Ignore nikoschopen |
9/14/2005 2:53:48 PM
Shel,
With regards to your previous filter, instead of using the days() qualifier, wouldn't it be better to use the count() qualifier? This way you can count how many times a given stock remained above or below an x indicator within the last 100 days. In addition, you've given 3 "if" statement, but left out a "then" clause of the conditional. Whether that is intentional or not, I would think you would be better served by adding a condition like "if 'the first if' is subtracted from 'the second if', show me the babes that are over 21" This way, it's easier to differential between the bull & the bear.
But here's what I find to be the most troublesome aspect about this type of filter. Suppose, for example, you get 4 days of "ema(13) is above ema(26)" & 4 days of "ema(13) is below ema(26)". You would think "hey, up 4 days followed by 4 down days, so it must go down" and vice versa. Even if the stock is above the MA, it doesn't necessarily imply that it's an up day; here I'm using it for simplicity sake) But upon closer scrutiny, you see that the stock is up 2 days, followed by 2 down days, followed by yet another 2 up days, only to tumble lower the next 2 days. In this case, you're looking at a range-bound stock, & you would think twice before betting your shirt with that hard earned money.
|
rtucker
318 posts
msg #38092
- Ignore rtucker
modified |
9/20/2005 3:51:10 AM
d
|
shelupinin
120 posts
msg #38094
- Ignore shelupinin |
9/20/2005 7:09:30 AM
and just another approach to rsi(3)<3 filter
P.S. Filter idea - my, filter coding - TheRumpedOne
Alex
|