nibor100 1,031 posts msg #149726 - Ignore nibor100 |
12/1/2019 2:20:19 PM
@xarlor,
I'm finally posting my attempt at a backtest filter for your Fibonacci screener, as I can't seem to make any more headway against SF complexity issues:
a. Before we get to that filter, here is my One day backtest filter, for your full Fibonacci screener, which shows which Fib Touch occurred the day before and how much the gain is, if one held from Open to Close the following day.
for example your Fibonacci Screener returns SAND for Tuesday Nov 26 and the One Day Backtest Filter returns SAND on Wed Nov 27.
b. The following filter conducts a 1 year backtest looking for the Fibonacci Screener Touch1 events.
It returns the stocks that have had winners or losers(next close to open % change) and the total hits for the past year for each stock. You'll see I've made many changes to your original filter in order to reduce the number of nested Set statements for complexity reasons but I believe it stays fairly close to your original Fib Screener resuts.
Note: It take at least 20 seconds to run each time.
c. If you run that filter in SF 2.0 you can hover at the apex of winners and losers triangles on their respective charts and quickly see the exact value of the gain or loss.
d. It can be fairly easily modified to do backtests for the Touch2, 3, 4, and 5 scenarios and you can adjust the date offset to go past 1 year in the past.
e. When I tried to run it with both Touch 1 and Touch2 criteria activated it kept running for at least 8 hours without returning results or error messages so i believe it can't run in SF.
Any help to expand the backtest to more Touch Fib lines would be appreciated or any other comments.
Ed S.
@Cheese,
I'm moving on the Weinstein backtest now in my spare time.....
|
Cheese 1,374 posts msg #149727 - Ignore Cheese |
12/1/2019 3:05:36 PM
Ed S nibor100 wrote:
@Cheese,
I'm moving on the Weinstein backtest now in my spare time....
=============================================================
Thank you very much for your time, Ed.
I am a basic sub so I had to make some minor editing.to your code.
Depending on the limit settings,
there were about 24 next day winners
and about 22 next day losers
out of 7,398 to 9,737 results on Nov 29, 2019
|
nibor100 1,031 posts msg #149729 - Ignore nibor100 |
12/1/2019 5:47:01 PM
@cheese,
please post your minor edited filter as I am very curious to what you did
thanks
Ed S
|
Cheese 1,374 posts msg #149730 - Ignore Cheese |
12/1/2019 6:29:55 PM
All I try to do here, as a basic sub, is see your backtest results with Nov 29, 2019 data.
sF still shows advanced sub required when I preview my post
but my edited version did run fine on my machine.
I do like some of graftonian's expanded works with Weinstein,
but the Weinstein basic specs didn't resonate with me.
Thanks again, Ed S nibor100.
|
nibor100 1,031 posts msg #149756 - Ignore nibor100 |
12/2/2019 1:12:28 PM
@All,
When I posted my backtest filter yesterday I posted an earlier test version by mistake, so the 200s for var5 and var8 should be replaced by 252s in order to cover a full year of data in the backtest.
Another OOps,
Ed S.
|
Cheese 1,374 posts msg #149759 - Ignore Cheese |
12/2/2019 1:54:50 PM
ok, thanks Ed.
|
nibor100 1,031 posts msg #149760 - Ignore nibor100 |
12/2/2019 2:08:23 PM
@cheese,
for Nov 29 my backtest filter returns 76 stocks vs your modified filter run on same date returning 40.
all 40 of your results stocks are in the group of 76.
My 76 have only 2 OTCBB exchange traded stocks,
My 76 have 29 stock with volume between 275,300 and 1,000,000
If I change your volume speed parameter to >275000 it still executes much quicker than my original, in about 6 seconds.
If I change my original var5 and var8 to just 100 days mine runs faster and returns 43 results for Nov29 vs your 40 results.
I thought your changes to make the filter faster were a good idea and I'm surprised that it mostly runs under the basic subscription when the addition of one more line makes mine un-runnable in my higher level subscription????
Thanks,
Ed S.
|
Cheese 1,374 posts msg #149761 - Ignore Cheese |
12/2/2019 2:30:50 PM
Thank you very much Ed.
On my machine, sometimes sF requires advanced sub, other times not.
I wonder if that may be due to the sF systems load at the time.
There were times when I was tempted to pay for an advanced sub.
But there is a benefit to be a basic sub. It forces me to look for the core of the filter.
I will have to take a closer look later.
|
Cheese 1,374 posts msg #149765 - Ignore Cheese modified |
12/2/2019 8:46:40 PM
Hello Ed,
ok, so I edited my basic sub version again to come closer to your adanced sub version.
This is what I got:
Dec 2, 2019 data
1,088 results
34 next day winners (33 instances)
37 next day losers (34 instances)
Nov 29, 2019 35
1,089 results
35 next day winners (34 instances)
37 next day losers (34 instances)
Based on counts, it was almost random, whis is what I would expect in many cases.
Thank you again for your time, Ed.
|
Cheese 1,374 posts msg #149766 - Ignore Cheese |
12/2/2019 8:51:39 PM
In case you may be interested, this is the edited version of Dec 2, 2019 in the evening.
Again, it worked on my machone but the preview suggests it requires advanced sub on sF machine.
|