comandante
5 posts
msg #32357
- Ignore comandante |
6/16/2004 5:36:38 AM
hi
i'm looking for a quick way to get all the stocks for which the pattern is candlestick bullish (or bearish). right now it seems i have to run the different patterns one at the time, quite a time consuming effort as there are at least 15 patterns to run.
of course the result will give hundreds of stocks, but i will refine my search with the use of indicators.
thanks
ps just signed in, sorry if the question is stupid...
|
cegis
235 posts
msg #32358
- Ignore cegis |
6/16/2004 10:56:05 AM
comandante,
Probably the easiest way is to set up a filter for each pattern, then set up a watch list and use auto-fill to fill the list with the results of all of the lists. Then, all you'd have to do is re-run the auto-fill on the watch list each night.
Another way - within a single filter - would be to use the count(<pattern>,1) function within set{} command to set variables to 1 (has pattern) or 0 (no pattern), then check to see of any of the variables are 1. I mention this because it could be "easily" combined with other technical indicators in a filter to get some very complex combinations.
HTH,
C
|
comandante
5 posts
msg #32361
- Ignore comandante |
6/16/2004 3:31:53 PM
thanks, that is exactly what i needed.
is there a maximum number of "set" commands that i can use per filter? i can't make it work with more than 17!
|
cegis
235 posts
msg #32362
- Ignore cegis |
6/16/2004 4:37:53 PM
There IS a limit to the levels of NESTING you can do. From what I understand, it has something to do with the LENGTH of the "unnested" set{} command exceeding some limit.
Try using set{} commands that are not defined based on other set{} commands.
If you post what you have (and the problem is indeed nesting), perhaps I can help with an alternate implementation...
C
|
comandante
5 posts
msg #32365
- Ignore comandante |
6/16/2004 5:15:41 PM
this is what i have, if i add one more set command i get 0 matches in return.
i wrote it based on a filter i found on another post
am i doing something wrong?
thanks
set{c1, count(pattern is bullish upside tasuki gap,1)} set{c2, count(pattern is bullish side-by-side white lines,1)} set{c3, count(pattern is bullish separating lines,1)} set{c4, count(pattern is bullish three line strike,1)} set{c5, count(pattern is bullish upside gap three methods,1)} set{c6, count(pattern is bullish harami cross,1)} set{c7, count(pattern is bullish three outside down,1)} set{c8, count(pattern is bullish three inside up,1)} set{c9, count(pattern is bullish homing pigeon,1)} set{c10, count(pattern is bullish bullish harami,1)} set{c11, count(pattern is bullish morning doji star,1)} set{c12, count(pattern is bullish tri-star,1)} set{c13, count(pattern is bullish meeting lines,1)} set{c14, count(pattern is bullish unique three rivers,1)} set{c15, count(pattern is bullish abandoned baby,1)} set{c16, count(pattern is bullish matching low,1)} set{c17, count(pattern is bullish engulfing,1)} set{OR1, c1 + c2} set{OR2, OR1 + c3} set{OR3, OR2 + c4} set{OR4, OR3 + c5} set{OR5, OR4 + c6} set{OR6, OR5 + c7} set{OR7, OR6 + c8} set{OR8, OR7 + c9} set{OR9, OR8 + c10} set{OR10, OR9 + c11} set{OR11, OR10 + c12} set{OR12, OR11 + c13} set{OR13, OR12 + c14} set{OR14, OR13 + c15} set{OR15, OR14 + c16} set{OR16, OR15 + c17} show stocks where OR16 is above 0
|
cegis
235 posts
msg #32373
- Ignore cegis |
6/17/2004 12:48:59 PM
comandante,
Yeah, this is the nesting problem. From what I've deduced (from various sources), the way SF interprets the set{} command for OR16 is that it replaces OR15 with it's definition, then OR14 with it's, and so on to get set{OR16,c1 + c2 + c3 ... + c16}, which it then substitutes all of the count()'s that defines each of the "c" variables to get the final calculation that's required to actually RUN the filter. Obviously, in this case, that's a pretty long string! (In fact, I'm surprised it can go 16 levels!)
I've been trying to figure out a way around this, so you can go further with this one filter, but alas, one escapes me (at the moment). I looked at using the algebraic fact that (A or B) is equivalent to NOT[ NOT(A) and NOT(B) ], so you could use the AND operator that SF handles so well, but I can't think of how to implement that final NOT of the whole condition.
One thing you could try is to add another group of set{} commands to add up another 16 condition variables (call the result OR16b), then have your filter condition (i.e., NOT in a set{} command) be "OR16 + OR16b is greater than 0". I'm not sure if that's gonna work, though. It will depend on if the limit is with the set{} command, or if it's with the total filter length. (This would get you to 32 conditions. I don't think you'd be able to go further...) You might be able to define c17 (just one more condition), and try "OR16 + c17 is greater than 0" to see if the limit is filter-wide. (If this method works, PLEASE post to let us know!)
HTH,
C
|
comandante
5 posts
msg #32377
- Ignore comandante |
6/18/2004 6:28:23 AM
cegis
i'll have a look at it in the weekend.
it is frustrating because 80% of my trading decisions come from simply watching candlestick charts.
i just need stockfetcher to help me pick stocks that show a good chart... and then i can START trading!
i don't want to spend too much time in creating a filter, because then i don't have any time to study the chart...
anyway thanks very much for your help, i'll let you know if i come up with some solution to the nesting problem
|
cegis
235 posts
msg #32379
- Ignore cegis |
6/18/2004 10:43:30 AM
comandante,
Don't forget my first post: Probably the easiest and quickest way is to set up filters for each individual pattern (or 16 at a time, if you prefer), then add the filters to an auto-fill watch list. There is hardly ANY difference to what you get for results (there are different columns, but you'll get the same stocks). If you want to spend minimal time, this would be the way I'd suggest.
HTH,
C
|