Cheese
1,374 posts
msg #152452
- Ignore Cheese |
6/4/2020 2:53:37 PM
Hope this is what you were looking for.
|
Cheese
1,374 posts
msg #152453
- Ignore Cheese |
6/4/2020 2:57:17 PM
|
styliten
303 posts
msg #152454
- Ignore styliten
modified |
6/4/2020 2:59:05 PM
Well, that's exactly what I meant.
If you round 0.10 with 1 decimal place you get 0.1
If you round 0.17 with 1 decimal place you get 0.2
In other words, if you attempt to DRAW Upper Standard Error Band(50) and alter the band width from 0.01 to 0.10 in 0.01 increment, you will notice that
Upper Standard Error Band(50, 0.01 thru 0.04) ALL appear the same because those values are rounded to 0; whereas:
Upper Standard Error Band(50, 0.05 thru 0.14 in 0.01 increment) are all the same thing as if Upper Standard Error Band(50, 0.10). 0.15 thru 0.249999 = 0.2; 0.25 thru 0.349999 = 0.3. So on and so forth.
Linear Regression Channels in comparison, does utilize 2 decimal places:
https://www.stockfetcher.com/forums/Indicators/Linear-Regression-Channels/43
so that:
draw Center Linear Regression Line(10, 0.01 thru 0.10 in 0.01 increment) would ALL have different appearance on the chart; the bigger the bandwidth, the wider the channel in appearance.
SF needs to at least make # of decimal places consistent across various linear regression related indicators. And as far as I am concerned, this change can not be considered technically difficult to do AT ALL. The fact there has never been any reasonable explanation not to implement or give my request a serious consideration really baffled me. I seriously wonder if there is still an active product development left at SF at this moment.
|
Cheese
1,374 posts
msg #152455
- Ignore Cheese |
6/4/2020 3:05:23 PM
This is what I got.
Maybe you can ask sF staff to clarify the level of rounding
|
Cheese
1,374 posts
msg #152456
- Ignore Cheese |
6/4/2020 3:07:36 PM
Maybe the display is rounded, but internally, the computing details are still there.
|
styliten
303 posts
msg #152457
- Ignore styliten |
6/4/2020 3:11:50 PM
Rounded only for the display could be true; at any rate, fixing that problem can not be a very difficult thing to accomplish at all. I've waited for more than 6 months for a reasonable response from SF's product development team but have got nothing at all back from them.
|
Cheese
1,374 posts
msg #152458
- Ignore Cheese |
6/4/2020 3:22:35 PM
No evidence and no expertise here, but
I believe sometimes Kevin_in_GA was able to communicate with sF staff in terms of some
lower level computing language and provide pseudo-code, which seemed to have facilitated
change implementation.
|
styliten
303 posts
msg #152459
- Ignore styliten
modified |
6/4/2020 3:34:04 PM
But Kevin_in_GA has left the building too! I even tried to initiate a contact with him from LinkedIn, but he is 3rd degree from me so I couldn't even send him an email from LinkedIn.
Here is Kevin's LinkedIn page:
https://www.linkedin.com/in/kevin-mcgrath-phd/
Kevin was at one time OBVIOUSLY interested in having a linear regression envelope developed. Apparently even he wasn't aware of the fact that Linear Regression Envelope had indeed already been in existence, not under that name but the much less intuitive name of "std error band"!
https://www.stockfetcher.com/forums/Filter-Exchange/Any-chance-of-seeing-a-LRI-Envelope/108033/-1/108034
That communication took place back in 2012; but SEB has been official since 2010!
https://www.stockfetcher.com/forums/Indicators/Standard-Error-Bands/96315
Further substantiating my earlier assertion that few people are comfortable with Linear Regression tools and even the SF staff/Tomb? who replied to Kevin's inquiry back in 2012 DIDN'T make the connection at all that Kevin's wish has already been realized! (albeit with the remnant problem of truncated/rounded display that you just nailed down!)
Life comes in full circles, doesn't it? What one doesn't know does make repeated comebacks to hurt us, right?
|
Cheese
1,374 posts
msg #152461
- Ignore Cheese |
6/4/2020 3:48:27 PM
I've just remembered that graftonian is also very good with complex coding tasks.
So, if people like xarlor, Ed S nibor100, mahkoh, graftonian, V E, snappyfrog and safetrade, etc
can custom build SEB from LRI and SE components,
then maybe you'll get more improved trading tools.
|
styliten
303 posts
msg #152462
- Ignore styliten
modified |
6/4/2020 4:02:10 PM
Kevin already gave a highly reasonable workaround in his original post:
draw lri(50) multiply by 0.998 on plot price
draw lri(50) multiply by 1.002 on plot price
I much prefer SF fix the display truncation/rounding problem because the built-in displays are also color-coded. The first SEB is colored RED, then GREEN, then PURPLE
Compare
Dow 30
Chart-time is 15
draw lri(10) multiply by 0.998 on plot price
draw lri(10) multiply by 1.002 on plot price
draw lri(25) multiply by 0.998 on plot price
draw lri(25) multiply by 1.002 on plot price
draw lri(50) multiply by 0.998 on plot price
draw lri(50) multiply by 1.002 on plot price
vs
Dow 30
chart-time is 15
draw center standard error band(10, 0.1)
draw center standard error band(25, 0.1)
draw center standard error band(50, 0.1)
Which one is visually more appealing to you? If you have to make a living viewing hundreds of them at a time and 10 to 15 times a day, which would you rather have? Not to mention the 2nd version is 3 lines shorter--I am now definitely in the "less is more" camp.
|