Download software Tutorial videos
Subscription & data-feed pricing Class schedule


New account application Trading resources
Margin rates Stock & option commissions

Attention: Discussion forums are read-only for extended maintenance until further notice.
Welcome Guest, please sign in to participate in a discussion. Search | Active Topics |

Using Linear Regression Sorts to Help Spot Divergences Topic Rating:
Previous Topic · Next Topic Watch this topic · Print this topic ·
Craig_S
Posted : Wednesday, March 9, 2005 4:14:23 PM


Worden Trainer

Joined: 10/1/2004
Posts: 18,819








If you prefer, you can download the video and watch it offline. Choose the Save option and save it to your hard drive. Extract the files to a location on your hard drive and double click on the .html file to launch the video. You must extract the files to your hard drive. You cannot play the video from the .zip file.

linear regression divergence sort sorting by price

- Craig
Here to Help!

To view this video you must have Macromedia Flash Player and Microsoft Internet Explorer 5+
Tanstaafl
Posted : Wednesday, March 9, 2005 5:38:00 PM
Registered User
Joined: 10/7/2004
Posts: 799
Location: Duluth, GA
Excellent video, Craig!

Some other nuances about divergence searches can be found in this thread:

Finding and Ranking Divergences, Visually

Craig_S
Posted : Wednesday, March 9, 2005 5:40:23 PM


Worden Trainer

Joined: 10/1/2004
Posts: 18,819
Thanks!

- Craig
Here to Help!
TopGun
Posted : Friday, March 11, 2005 8:08:00 PM
Registered User
Joined: 3/11/2005
Posts: 4
LINEAR REGRESSION DIVERGENCE SORT, EXCELLENT VIDEO.

THANK YOU
Doug_H
Posted : Friday, March 11, 2005 8:09:44 PM


Worden Trainer

Joined: 10/1/2004
Posts: 4,308
Glad you liked it, TOPGUN!

- Doug
Teaching Online!
Darkrider
Posted : Monday, September 26, 2005 3:59:29 PM
Registered User
Joined: 11/21/2004
Posts: 13
Is there a way to build price/TSV regressions as PCFs?
Craig_S
Posted : Monday, September 26, 2005 4:03:02 PM


Worden Trainer

Joined: 10/1/2004
Posts: 18,819
What periods would you like the LRs and what period is the TSV?

- Craig
Here to Help!
Darkrider
Posted : Monday, September 26, 2005 4:34:10 PM
Registered User
Joined: 11/21/2004
Posts: 13
More days are better on any regression, however given that I'm looking for divergences, I'd suspect that another over 50 days would probably be less useful. As far as TSV periodicity, I'm not sure what would be appropriate. I'm still new to the TSV calculations so if there's a min/max range that is necessary, I'd work within that. My gut says use 21-day TSV, but I'm very much open to suggestions.
Craig_S
Posted : Monday, September 26, 2005 4:45:43 PM


Worden Trainer

Joined: 10/1/2004
Posts: 18,819
They are pretty hefty formulas.

Do me a favor. Play with the method taught here until you find a TSV and LR period you love. Come back and I will be happy to try and put something together for you.

Is that fair?

One thing that is nice about this method is you can change your settings on the fly without creating and calculating long formulas each time.

- Craig
Here to Help!
Darkrider
Posted : Wednesday, September 28, 2005 12:05:45 AM
Registered User
Joined: 11/21/2004
Posts: 13
Is there a way to send you a private message? I've played w. the video instruction and set up a couple of PCFs but would like to discuss w. you off-forum if possible. TIA
Craig_S
Posted : Wednesday, September 28, 2005 3:54:40 AM


Worden Trainer

Joined: 10/1/2004
Posts: 18,819
You are welcome to send me an email. Email support@worden.com and put ATTN: Craig Shipman in the subject line.

- Craig
Here to Help!
Darkrider
Posted : Thursday, September 29, 2005 7:10:58 AM
Registered User
Joined: 11/21/2004
Posts: 13
Craig,

Here are the formulas I'm using. I've decided to calculate 10-day slopes for both price and TSV and have set up PCF for price slope as follows:

((((C + (C1 * 2) + (C2 * 3) + (C3 * 4) + (C4 * 5) + (C5 * 6) + (C6 * 7) + (C7 * 8) + (C8 * 9) + (C9 * 10)) * 10) - (55 * (C + C1 + C2 + C3 + C4 + C5 + C6 + C7 + C8 + C9))) / ((((1 * 1) + (2 * 2) + (3 * 3) + (4 * 4) + (5 * 5) + (6 * 6) + (7 * 7) + (8 * 8) + (9 * 9) + (10 * 10)) * 10) - (55 * 55)))

I then took this formula and added a subtraction of the TSV slope as follows:

((((TSV10 + (TSV10.1 * 2) + (TSV10.2 * 3) + (TSV10.3 * 4) + (TSV10.4 * 5) + (TSV10.5 * 6) + (TSV10.6 * 7) + (TSV10.7 * 8) + (TSV10.8 * 9) + (TSV10.9 * 10)) * 10) - (55 * (TSV10 + TSV10.1 + TSV10.2 + TSV10.3 + TSV10.4 + TSV10.5 + TSV10.6 + TSV10.7 + TSV10.8 + TSV10.9))) / ((((1 * 1) + (2 * 2) + (3 * 3) + (4 * 4) + (5 * 5) + (6 * 6) + (7 * 7) + (8 * 8) + (9 * 9) + (10 * 10)) * 10) - (55 * 55)))

When I combine the 2, I should get a solution of the slope of price less the slope of TSV:

((((C + (C1 * 2) + (C2 * 3) + (C3 * 4) + (C4 * 5) + (C5 * 6) + (C6 * 7) + (C7 * 8) + (C8 * 9) + (C9 * 10)) * 10) - (55 * (C + C1 + C2 + C3 + C4 + C5 + C6 + C7 + C8 + C9))) / ((((1 * 1) + (2 * 2) + (3 * 3) + (4 * 4) + (5 * 5) + (6 * 6) + (7 * 7) + (8 * 8) + (9 * 9) + (10 * 10)) * 10) - (55 * 55))) - ((((TSV10 + (TSV10.1 * 2) + (TSV10.2 * 3) + (TSV10.3 * 4) + (TSV10.4 * 5) + (TSV10.5 * 6) + (TSV10.6 * 7) + (TSV10.7 * 8) + (TSV10.8 * 9) + (TSV10.9 * 10)) * 10) - (55 * (TSV10 + TSV10.1 + TSV10.2 + TSV10.3 + TSV10.4 + TSV10.5 + TSV10.6 + TSV10.7 + TSV10.8 + TSV10.9))) / ((((1 * 1) + (2 * 2) + (3 * 3) + (4 * 4) + (5 * 5) + (6 * 6) + (7 * 7) + (8 * 8) + (9 * 9) + (10 * 10)) * 10) - (55 * 55)))

The problem I'm having is that I want to find companies where the 2 metrics' slopes are diverging, so I need to look at the change from yesterday to today. I've calculated that PCF as well, but started to realize that unless I can get them into an Easyscan, I'm wasting my time. What I'm attempting to do is identify on at least a daily basis (I'd love to do real time, perhaps 2x/day), companies where the relationship between price and TSV is improving (long) or eroding (short). I can add other variables to get the list smaller if I end up w. too many companies, but I'm really interesting in monitoring closely those names were TSV is doing something opposite of price.

Any suggestions you can provide regarding creating this type of model into an Easyscan would be greatly appreciated.

Craig_S
Posted : Thursday, September 29, 2005 8:00:07 AM


Worden Trainer

Joined: 10/1/2004
Posts: 18,819
Here is how I would do it.

Create the following PCF:

(AVGC5-AVGC5.5)*(AVG(TSV10,5)-AVG(TSV10.5,5))

When this returns a POSITIVE number, the price LR and TSV LR are "in step" with each other. WHen this returns a NEGATIVE number, the two are diverging (one sloped positive, one negative).

Create the PCF and then sort some WatchLists by them. You will get a clear sense between the values it returns and what you are seeing on the chart.

I am simplifying things greatly here but I bet you will find this more than gets the job done. Try it!

Once you get a sense of the values-to-chart relationship by sorting, you can easily limit the values when you use it in EasyScan.

- Craig
Here to Help!
Darkrider
Posted : Friday, September 30, 2005 8:43:32 AM
Registered User
Joined: 11/21/2004
Posts: 13
It seems to work well in sorts and turns up some interesting charts in EasyScan. My next question regards your thoughts regarding "optimal" periodicity (as if there is such a thing). Obviously with data, more is better, however I'm looking to get a sense for the short-term flow of funds that could signal institutional movement. Is there a minimal # of data points below which you'd conclude that the TSV LS is meaningless? I'm asking b/c you've seen a lot more of the raw TSV calcs than I ever will and I'm attempting to determine if 10 days is sufficient to get a valid TSV calc. Perhaps to re-ask the question, what's the mininum # of days required to calculate a TSV that is valid (I'm assuming it's probably more than 2 and less than 100). TIA
Bruce_L
Posted : Friday, September 30, 2005 9:50:27 AM


Worden Trainer

Joined: 10/7/2004
Posts: 65,138
Darkrider,
I'm not in a position to give you "optimal" settings for TSV (or even a minimum period). The following forum topic may be of interest:

Peter Worden's TSV

-Bruce
Personal Criteria Formulas
TC2000 Support Articles
Darkrider
Posted : Friday, September 30, 2005 10:22:23 AM
Registered User
Joined: 11/21/2004
Posts: 13
Allow me to rephrase the question then...does a TSV1 provide the same data regarding money flows as a TSV10? Perhaps if you can provide a link to the TSV calc, I can judge for myself if you'd prefer. Thx
Bruce_L
Posted : Friday, September 30, 2005 10:27:24 AM


Worden Trainer

Joined: 10/7/2004
Posts: 65,138
The calculation of Time Segmented Volume is proprietary, so I can’t provide the formula, but TSV10 is a 10-Period Simple Moving Average of TSV1. The following videos should help you judge for yourself how TSV1 relates to TSV10:

Understanding Moving Averages - Part I, The Basics
Understanding Moving Averages - Part II, Using Multiple Averages

-Bruce
Personal Criteria Formulas
TC2000 Support Articles
Darkrider
Posted : Friday, September 30, 2005 3:44:05 PM
Registered User
Joined: 11/21/2004
Posts: 13
Thanks. So if I'm comparing a 10-day LR of price and want a comparable LR of TSV, I should merely use TSV rather than TSV10 correct? By doing it my original way (10day LR of price vs 10day LR of TSV10), I'm using 10 days of price data and 20 days of TSV data, isn't that correct?

Bruce_L
Posted : Friday, September 30, 2005 4:16:10 PM


Worden Trainer

Joined: 10/7/2004
Posts: 65,138
Darkrider,
You are a little off on the number of days when doing 10-Period Linear Regressions of Price (10 days) and TSV10 (19 days of TSV1, although it might require more days of data). I'm not sure if comparing the same number of days is all that important, but as a Trainer I can't tell you what settings to use or how to interpret an indicator.

-Bruce
Personal Criteria Formulas
TC2000 Support Articles
bustermu
Posted : Saturday, October 8, 2005 10:40:59 AM
Registered User
Joined: 1/1/2005
Posts: 2,645
QUOTE (Darkrider)
Allow me to rephrase the question then...does a TSV1 provide the same data regarding money flows as a TSV10? Perhaps if you can provide a link to the TSV calc, I can judge for myself if you'd prefer. Thx


Darkrider,

Bruce has already stated that TSV10 = AVG(TSV1,10).

It may be helpful for you to know that the values of

(TSV1.0,TSV1.1,TSV1.2,TSV1.3,TSV1.4,TSV1.5,TSV1.6,TSV1.7,TSV1.8)

convey no information whatsoever about the values of

(TSV10.0,TSV10.1,TSV10.2,TSV10.3,TSV10.4,TSV10.5,TSV10.6,TSV10.7,TSV10.8)

and conversely. We can construct charts for which each of the above two sequences can assume any two preassigned sequence values whatever.

Please notice that we stopped the sequences at 9 values, just short of the period of TSV10, in order to make this statement. This is a property of Simple Moving Averages and has nothing to do with TSV per se. Please see:

A Property of Simple Moving Averages

TSV10 Simple is superb at telling you whether there is Net Share Accumulation or Distribution, as measured by TSV, over any interval whose length is a multiple of 10. It cannot with certainty convey such information for any other interval lengths. For example, if you want to know whether there is Net Share Accumulation over some interval of length 15, TSV1, TSV3 Simple, TSV5 Simple, or TSV15 Simple can tell you, but no other TSV Simple can including TSV10 Simple.

If Craig had illustrated his fine approach with MS instead of TSV these problems would not have arisen here.

Thanks,
Jim Murphy
albeshi
Posted : Friday, October 12, 2007 11:49:13 PM
Registered User
Joined: 9/28/2006
Posts: 1
thank all >>
funnymony
Posted : Monday, May 5, 2008 11:48:40 AM

Registered User
Joined: 2/5/2006
Posts: 1,148
QUOTE (Craig_S)
Here is how I would do it.

Create the following PCF:

(AVGC5-AVGC5.5)*(AVG(TSV10,5)-AVG(TSV10.5,5))

When this returns a POSITIVE number, the price LR and TSV LR are "in step" with each other. WHen this returns a NEGATIVE number, the two are diverging (one sloped positive, one negative).

Create the PCF and then sort some WatchLists by them. You will get a clear sense between the values it returns and what you are seeing on the chart.

I am simplifying things greatly here but I bet you will find this more than gets the job done. Try it!

Once you get a sense of the values-to-chart relationship by sorting, you can easily limit the values when you use it in EasyScan.



i assume the formula above is for a 10 day period. if i want change the formula to analyze 25 days do i just change to 10's to 25's? not really sure what the 5's represent in your formula.
malinbeg
Posted : Saturday, March 28, 2009 2:36:32 PM
Registered User
Joined: 3/10/2005
Posts: 3
How can I save training video to my computer
Bruce_L
Posted : Monday, March 30, 2009 11:27:30 AM


Worden Trainer

Joined: 10/7/2004
Posts: 65,138
malinbeg,
We've added an option to download the video to the original post.

-Bruce
Personal Criteria Formulas
TC2000 Support Articles
mussie2
Posted : Wednesday, July 22, 2009 8:59:39 PM
Registered User
Joined: 10/31/2004
Posts: 12
Location: deerfeild breach Fl.
QUOTE (Craig_S)
Here is how I would do it.

Create the following PCF:

(AVGC5-AVGC5.5)*(AVG(TSV10,5)-AVG(TSV10.5,5))

When this returns a POSITIVE number, the price LR and TSV LR are "in step" with each other. WHen this returns a NEGATIVE number, the two are diverging (one sloped positive, one negative).

Create the PCF and then sort some WatchLists by them. You will get a clear sense between the values it returns and what you are seeing on the chart.

I am simplifying things greatly here but I bet you will find this more than gets the job done. Try it!

Once you get a sense of the values-to-chart relationship by sorting, you can easily limit the values when you use it in EasyScan.
mussie2
Posted : Wednesday, July 22, 2009 9:02:59 PM
Registered User
Joined: 10/31/2004
Posts: 12
Location: deerfeild breach Fl.
I copied the PCF but how do I test it? or what do I do next?   mussie
Bruce_L
Posted : Thursday, July 23, 2009 9:15:50 AM


Worden Trainer

Joined: 10/7/2004
Posts: 65,138
mussie,
You would test the Personal Criteria Formula by selecting Test in the Editing Formula window for the PCF. The How to create a Personal Criteria Forumula (PCF) covers this in more detail. Craig suggested sorting by the PCF. The Create your own stock rankings using WatchLists and Sort conditions topic explains how to do so.

-Bruce
Personal Criteria Formulas
TC2000 Support Articles
Users browsing this topic
Guest-1

Forum Jump
You cannot post new topics in this forum.
You cannot reply to topics in this forum.
You cannot delete your posts in this forum.
You cannot edit your posts in this forum.
You cannot create polls in this forum.
You cannot vote in polls in this forum.