fotchstecker
304 posts
msg #154500
- Ignore fotchstecker
modified |
10/29/2020 6:53:41 PM
Nibor, I could hardly ask you to spend your time trying, but this is the filter I had in mind. It's just a very slightly modified version of the "catalyst" filter.
Take for example TCON, which I think last appears on Sep 8. However, if you run the filter back in time from there, you will see TCON appear a number of times, and it has a significant runup in the prior weeks from its first appearance within the last three months. So, it would be useful to see, when running a filter on the latest date, if and how many times a symbol has appeared over the last X period. This is useful for a variety of reasons -- like knowing if a symbol is a "newcomer" or if a symbol is on a long stretch of appearances. It could be useful to get a full percent-change in price stat for the symbol, too, over the period. I have the enhanced/full SF subscription.
Anway, I don't expect you to try to work this out at all, but I did want to reply to share more of what I was thinking. Getting improved stats around a symbol in general would increase the power of SF a lot.
I would also add that SF could improve the ability to rank symbols and to track changes in rank -- like used in rotation models. But that's another thread for another time.
|
nibor100
1,046 posts
msg #154511
- Ignore nibor100 |
10/31/2020 2:22:38 AM
@fotchstecker,
1. This was a very easy task as Xarlor had updated the "catalyst filter" back on May 11th in the "This might be impossible" thread with these 2 very important backtest lines:
count(buy > 0,251) > 0
count(buy 1 day ago equals 0,251) > 0
However, you must follow his instructions in the comment lines, to the letter, to get the historical results.
2, To see the backtest statistics for the past year in a column display change the first line above to
add column count(buy > 0,251) > 0
3. To get them sorted by most to least add this other line
sort on column 5 descending
Ed S.
|
Cheese
1,374 posts
msg #154512
- Ignore Cheese |
10/31/2020 12:19:45 PM
@nibor100
I'm impressed. You come up with solution so quickly,
@fotchstecker
If snappyfrog's DRAW CMA(PVI,22) is appended to Xarlor's filter with Ed's suggested solution,
then CMA(PVI,22) could show the way to a big opportunity for your TCON example around Jan 2019.
Just a thought.
|
fotchstecker
304 posts
msg #154517
- Ignore fotchstecker |
11/1/2020 1:34:01 PM
@nibor
How did I miss this?! Thank you so much for pointing it out for me -- it's exactly the thing I was wanting to add. Thank you, Nibor and Xarlor.
|
fotchstecker
304 posts
msg #154518
- Ignore fotchstecker |
11/1/2020 1:35:08 PM
Cheese -- you are a step ahead of me. I was thinking of adding the same thing but perhaps in a different way. I think your idea is better and will try snappy's code before testing mine. Thank you.
|
Cheese
1,374 posts
msg #154519
- Ignore Cheese |
11/1/2020 1:50:05 PM
@fotchstecker
You may also want to look at TCON and your other picks with snappyfrog's newer filter(s) here
https://www.stockfetcher.com/forums/Filter-Exchange/Weekly-and-Daily-MACD-on-Same-Graph/154365
You may want to use snappyfrog's filters to cast a bigger net for potential gains,
and use the Xarlor's filter that you mentioned above as confirmer in some cases.
It seems to me that there is synergy between the three moving averages in Xalor's filter
and the signals in the aforementioned snappyfrog's
Happy trading
|
nibor100
1,046 posts
msg #154537
- Ignore nibor100 |
11/3/2020 2:06:48 PM
@fotchstecker,
When I run your filter, modified for backtest hit statistics it show only 2 hits in the last year for TCON, both are in September.
If I add the following line:
date offset is 11/3/2019
there are no TCON hits shown for the year prior to that date.
the same is true for each of these date replacements:
date offset is 11/3/2018
date offset is 11/3/2017
date offset is 11/14/2016
So I don't quite follow your post of 10/29 where you stated:
"which I think last appears on Sep 8. However, if you run the filter back in time from there, you will see TCON appear a number of times,"
Could you provide me some of those dates that this filter found TCON or maybe it was a different version of the filter?
Thanks,
Ed S.
|
fotchstecker
304 posts
msg #154539
- Ignore fotchstecker
modified |
11/3/2020 5:54:57 PM
@nibor
You are correct and you are experiencing the same issue I noticed in the last day or so. To explain the sequence, I used TCON as an example a couple of posts ago because I had recalled running the filter and noticed it occurring several (at least 3) times. At the time, I noted the percentage changes from day to day and had a significant runup. So, remembering this, I wrote my post here and used it as an example, however, I did not double-check at the time whether the same output would occur; I assumed it would.
However, the version of the filter I posted does as you said -- I can find only 2 TCON appearances now, like you. I'm not certain if I changed anything, but something must be different. I will try to modify the lines I typically shift around minimally to see if I can reproduce the original result.
I'll update here with any notable findings. Thanks for checking through things, nibor.
|
fotchstecker
304 posts
msg #154540
- Ignore fotchstecker |
11/3/2020 6:43:22 PM
@nibor (reply 2)
Incidentally, I don't think I had to comment-out and swap-in the lines as explained in the revised version. I think the following seems to work to get both the count and the current signal. Do you agree it makes sense?
|
nibor100
1,046 posts
msg #154545
- Ignore nibor100 |
11/4/2020 12:49:32 PM
If all you want are the latest day's hits and their historical counts for the past year, your new version will work fine.
Ed S.
|