Search | User Listing Vidsite | Calendars | Quotes
Home Page ->  Custom TradeTight Routines -> Tools In Team Development -> View ThreadLogon (or Register, or Join TradeTight)

You are logged in as a limited-access Guest.To join TradeTight, first read the info in the Organization & Content room, then click the link above. 

James 1:5 (NIV) ... If any of you lacks wisdom, he should ask God, who gives generously to all without finding fault, and it will be given to him.


Sticky CVW Journal & DISCUSSION
Jump to page : 1
Now viewing page 1 [50 msgs/pg]
Jump to room :
JimDean
Posted 6/27/2019 11:20 AM (#9641)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
This thread "ongoing" Discussion, the best starting-point for those who joined the Team in June ... this is where I keep an ongoing "journal" of work, and post questions for team-members to help out with. This is the thread where your answers, comments and questions should be posted.

Other threads related to CVW development:

1. An active thread with "reference" posts such as help screens that will be kept up to date as development progresses (not open for user-posts):

2. An inactive thread with the initial VolEval Beta and SBAS testing (frozen now ... no more posts): http://tradetight.org/forums/thread-view.asp?tid=1431

3. An inactive (frozen) past-journal/discussion thread with CVW-development info, *prior* to adding several new Team members in June. It grew VERY long (5 pages), so I moved info from it to the VolEval thread and the Reference thread. There is a lot of interesting info about the thinking behind nearly every aspect of the "middle" phase of development, but most of those posts were later superseded by refined approaches ... so much of the info there does not accurately reflect how CVW operates now:
http://tradetight.org/forums/thread-view.asp?tid=1432


Important ... If you have comments or questions about a post in another (frozen) thread, then you MUST include a URL link in your post here, so that others can quickly find out what you're referring to. This is how to create that link (this is much easier to do than to explain, btw, and works for all forums, even N):

1. Navigate to the post in this Reference thread that you are mentioning.

2. COPY the entire URL string from the browser window ... for this post, it looks like this:
http://tradetight.org/forums/thread-view.asp?tid=1444&posts=1&start...
(actually, as this thread grows, the #posts and start# will change, or, depending on how you navigated to the reference-posts, the info after the "tid=xxxx" might be different)

3. Memorize the date-time info-line just above that post, for the "#nnnn" number that identifies that post. For this post, "Posted 6/27/2019 9:07 AM (#9641) ", what you want is: 9641

4. Navigate to this Discussion thread, and in your new post, PASTE the URL from step 2, and DELETE everything in the URL after the "tid=1444".

5. TYPE in a pound-sign followed "M" and the ID number of the post being referenced (step 3)

6. When clicked (or pasted in browser field), the completed URL should jump you directly to the post in question, and should end up looking like this:
http://tradetight.org/forums/thread-view.asp?tid=1444#M9641

NOTE: if you simply provide the #Mxxxx number, it's NOT a convenient means of finding the referenced post (requires you to use Search and potentially wade through things). So, please ALWAYS use this method.
Top of the page Bottom of the page
JimDean
Posted 6/27/2019 11:50 AM (#9643 - in reply to #9641)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
Here we go ...

Some of the more recent posts in the (now-archived) earlier discussions thread were moved directly into the Reference thread. However, there are a LOT of posts in that earlier 5-page 250-post thread that have info which is still relevant and current ... intermixed with stuff that's not really "reference" info.

So, I'm creating some new posts in the Reference thread that duplicate the info from posts that still remain in the archived one. I'll keep the ones in the Reference thread up to date, so eventually the archive thread copies will become invalid.
Top of the page Bottom of the page
JimDean
Posted 6/29/2019 9:51 AM (#9649 - in reply to #9643)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
Hi, all ... please feel free to post any comments or suggestions about this ... it's a fairly simple change, but I figured the explanation behind it would help y'all come up to speed.

I just updated the Reference snapshots of the "draft" versions of Help Panels #3 - #6, here:
http://tradetight.org/forums/thread-view.asp?tid=1445#M9634

Yesterday I finalized the color map for the WaveBand ... this is a crucial Stradicator paradigm output feature. I *updated it today*, adding "*" flags on NINE of the Bull|End|Bear labels. I posted a "crib sheet" color-key in the Reference thread, here:
http://tradetight.org/forums/thread-view.asp?tid=1445#M9647

This morning, I posted six sample charts in the Reference thread, that show the current status of the various Plot Format options. One of them illustrates all the "envelopes and dots" turned on (negative slider Powr|Cstm versions), and the other five show positive-slider all-version defaults. The Equity option (#3) will be fleshed out soon ... I need to finish up the envelope coding first. Check it out here:
http://tradetight.org/forums/thread-view.asp?tid=1445#M9648

Quick notes:
... the final Option #5 "V, CumV, MAs" plot is intended primarily as an "educational" output, providing a "drill-down view" of the Raw & Capped volume, CumVol Increm & Total, 3-MA's, and Slope & Stack status "all in one". It is NOT intended for actual "use" or decision-making, but rather as a "resource" when someone wants to know "why" the curves and histos and strips in Option #4 Vote Detail are what they are. This plot-option might only be available with the Cstm-version.
... the Option #4 Vote Detail is also an "explanatory" plot ... it shows the seven individual "votes" for Slope, CurBar, and Stack in a normalized form. Most people won't need this, since the three Option #2 bottom Strips show the "net" of those three categories ... but again, for people who are curious what's going on, this is very useful. This plot-option might only be available with Powr+Cstm versions.

The first sample show the default Plot Option #2 can get very "busy" on the PricePane, with all the envelopes turned on by the neg-slider BgnVot, EndVot, FuzVot settings (fewer envelopes if any/all are positive). Some folks like to see the "general reasons" behind actions taken by the Stradicator ... this combo of Price & Indicator panes is designed for that ... it's three thin Strips below the Band in the Indicator pane "explain" the status of the internal Slope, CurBar, and Stack voting, which collectively are the primary controlling factor for the Band.

However, some want a cleaner chart, and don't care too much about "reasons" ... they just want to "know what to do". Or, maybe they also want to plot some other things on the Price Pane, &/or create additional stacked Indicator panes for their own reasons. In that case, the "envelopes" may make the PP too "busy" ... and the "explanatory detail" from three bottom horizontal "strips" might not be needed. The second snapshot (Plot Option #1) shows that "clean action" mode ... the PricePane dots which indicate signals for Orders remain, as does the Band which echoes that info (plus general trade "status"), but the extraneous busy-ness is removed. (Note the PricePane dots all are echoed just below the Band.)

At this point, please review the first-linked M9647 Band-color reference-key. The snapshot has been updated, per the new fifth paragraph (just above the italics). The new asterisks flag nine "events" that logically only have to do with *managing* a real Trade: Price crossing Envelope lines (partial or full exits), Add-Ons, and Full exits related to position size or equity-vote filtering.

My "background" thinking has always been that the Stradicator "state" might be useful as a FILTER for other Strategies (or other Stradicators, or the same Stradicator with different settings ... think a bit on THAT one ;~). Presuming that's true, I'm thinking it might make sense to decouple some of the "trading-actions" logic for the "filter-mode" ... retaining the features that relate to the core of the Stradicator's focus (in this case, Cumulative Volume), and REMOVING the features related to raw price action (envelopes), or to position-size and equity-filtering.

If that kind of "filtering mode" is valuable/useful, then MAYBE the Plot (and Return) Output options could expand to include it. That is, a SIXTH plot option could be offered (it would become #1, and the current #1-#5 would become #2-#6) ... the new Filter-mode option #1 would have the seven "*" features inactive (which could alter the appearance of the Band ... sometimes allowing the "Waves" to continue longer than they would otherwise). The PricePane Plot would be completely clean ... no dots or envelopes. The Indicator Pane for the Filter mode would show the (altered) Band ... but nothing else (therefore using very little "chart real estate").

However ... without adding a new special "Filter" mode output option, it's possible to turn off all seven of those "*" features, with by making all the VoteGuru+Expert and CutGrab and BarWndo slider-settings POSITIVE, so that if you want to actually *trade* on that basis, you can. But if you're trading rather than filtering, my guess is that you'd want to see (at least) the dots on the price-pane, (possibly) some of the other plot-output options, and definitely many of the Return-value options (Stats for Scans, Signals for trades, etc).

In order to Keep It Simple & Sensible, I'd rather not add more output options ... so, this is what I'm thinking about doing ...

MODIFY Plot Option #1 so that it turns OFF its current PricePane dots IF all the VoteGuru+Expert and CutGrab+BarWndo sliders are all Positive (neg inputs are not avail for Easy|Xprt versions, btw).
That is, though "entries and exits" are shown by their related Band colors on the Indicator, Option #1 keeps the PricePane totally clean and the Indicator Pane simple. However, if any of those sliders are negative, you'd see the impact of the negative-input features on the Band with the #1 output, AND the Dots would show up on the Price-Pane (but no envelopes).
Using this approach, if you want to see explanatory detail ... any active PricePane envelopes plus 3-strips, or the Equity Status, or the Vote-Detail, or the V+CumV+MAs), you'd flip to plot Option #2,3,4,5

QUESTION ... do you think I should create new Plot/Return options just for the Filter mode, or do you think the "Modification" above with the current options is sufficient?

And ... this is a good time for you to take a look at that Band-Colors "key" as well as the five samples, and ask any "burning" questions you might have about the general format or approximate meanings.
Top of the page Bottom of the page
MikeDeWalt
Posted 6/29/2019 2:26 PM (#9651 - in reply to #9650)
Subject: CVW Journal & DISCUSSION



Spectator

Posts: 1
0
Location:
USA: WA, Stanwood
1. I believe than any internal logic that results in potential state change in orders or holding should also have some external manifestation. On the other hand internal states that generate signals and do not effect external behaviors can be masked, not displayed, or hidded and available via sliders or other options.
2 & 3 I have no opinion on colors other than the consistency issue that you've addressed in another post.
Top of the page Bottom of the page
JimDean
Posted 6/29/2019 3:00 PM (#9650 - in reply to #9649)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
Second Journal + Question post for 6/29 ... this one needs your replies ASAP ... you can wait on reviewing the prior post till after this one ;!)

If you've taken a reasonably careful look at the "BgnVotNetNatcMny2Fw" explanation on the Reference-thread Crib Sheet or Help Panel #4, then you'll recall that slider-positions of (pos or neg) 1,3,5 use "ToeIn + Confirm" entry logic, while slider = 2 or 4 use "Norm-Entry" (All-In) logic. Check out the explanation here:
http://tradetight.org/forums/thread-view.asp?tid=1445#M9625

Important background:

This choice is a "major feature" of Stradicators, supported by both the Simple and Robust Trade Plans. The first three "Bull Alert", "Bull Toe-In", and "Bull Confirm" bluish Band colors appear only if the BgnVot slider is 1, 3, 5 ... (also, the last-three "Bear" purplish Band colors). OTOH, if the slider is 2 or 4, then "Norm-Entry" starts every trade with a single large order (lawngreen or hotpink color). See the Color-Key here:
http://tradetight.org/forums/thread-view.asp?tid=1445#M9647

Once a Trade has made a Norm-Entry, or has had a Confirm order, it's into the "full-Wave" mode where the greenish or reddish Band colors take over ... that is, the trade is no longer in a "tentative" Alert aka ToeIn state (which has a smaller position size). ToeIn orders appear when the NetVote is one vote less than the required BgnVote (6,6,5,5,4 votes per BgnVot slider).

Sometimes the extra NetVote required to Confirm (or create a NormEntry) never appears ... in that case, the trade starts at a fraction (1/3-1/2) of the Confirm'd size, and never gets any bigger.
If the NetVote jumps straight to the BgnVot threshold without an Alert-Vote bar preceding it, then if the slider=1,3,5 Toe+Confirm, the *first* bar w/adequate NetVote is a ToeIn-sized order.
In either case, if another bar appears soon thereafter that meets the BgnVot threshold, *without* any "weakening" of the vote since the ToeIn before that, that Confirm >= BgnVot bar bumps the trade size up and the Wave turns green/red.

In green/red Wave mode, if the FuzVotBndAddsSm2Lrg is negative, then Add-On's can appear (there are important pre-qualifiers for them, though). Add-On's *never* can appear in "Toe-In" size/mode. Also, the green/red Wave mode has a sophisticated chunk code that creates "warnings" (dark red/green bars) from several different situations (controlled by the EndVotOpXmwFw2Mny slider) ... much of that logic just doesn't make sense for ToeIn mode.

In *either* the green/red Wave mode, or the blue/purple ToeIn mode, there are a large variety of situations where a Full Exit is signalled ... too many to discuss here ... and most of those rules apply in *both* modes. If Partial Cuts &/or Partial Grabs are activated, via negative CutGrabGapSzSm2Lrg input (showing the dim inner-envelope pricepane lines), then in *either* mode, Partial exits will occur when Close crosses those lines.

Currently, when in ToeIn (aka Alert) bluish/purplish mode, some of the FullExit rules aren't used ... instead of creating (two different ways) and counting "Warnings", then signaling a FullExit when enough appear, the ToeIn mode simply checks if the NetVote has dropped >= 2 since the ToeIn started ... if so, the ToeIn/Alert mode does a FullExit. Currently, there are no ToeIn-mode "Warnings" (potentially creating Partial Exits) derived from Slope+CurBar+Stack Voting.

PROPOSED NEW TOEIN-MODE WARNINGS:

Since the Partial CutGrab envelopes CAN currently do a partial-exit during ToeIn mode ... and since in the red/green Wave mode any PartialCut is treated as a Warning re the internal logic ... then it seems to me that there should be some kind of internal logic that permits a Warning (PartialExit) to fire during a bluish/purplish ToeIn mode ... that is, scaling out of the ToeIn size, rather than waiting for a Full Exit.

Since a Vote-logic ToeIn Full Exit fires when the NetVote drops by 2, it seems sensible to make the Vote-logic Warning (PXit) fire when the NetVote *drops by 1* below the initial ToeIn/Alert bar's vote.
Depending on the EndVotOpXmwFw2Mny slider setting, which defines the PX size (40%, 30%, or 20%) ... along with the BgnVotNetNatcMny2Fw slider setting, which defines the ToeIn size (60%, 40%, or 20%) ... it *could* be that *just one* Warning/PXit that's >= the size of the ToeIn, could automatically convert to a Full Exit, ending the ToeIn mode trade.
OTOH, if the ToeIn is 60% and the PX is 20%, then it might take *up to 3 Warnings* to fully scale-out of the ToeIn mode ... this presumes the NetVote drops by one, then goes up by one, then drops by one again, then goes up by one, then drops by one a third time ... pretty unlikely ... but the trade is "managed" properly ... the price is "chattering", so the size is gradually scaled back and finally the position is closed (if some other FullXit logic doesn't intervene first, which is more likely).

It's not difficult to implement this ToeIn mode refinement *at this point*. I could use Teal or DarkCyan for the Bull ToeIn Warning color, and Magenta or Orchid as the Bear ToeIn Warning color, bringing the total #Band colors to a well-arranged total of 29!
OR ... I could use the already-existing three greenish and three reddish Warning colors, which distinguish the reason creating the logic (and the Partial-Grab cross colors)

THREE QUESTIONS:

1. Do you think I should implement the "ToeIn Mode Warning" logic as described above? If not, WHY not?

2. After checking the full-palate of colors, here:
http://tradetight.org/forums/thread-view.asp?tid=1445#M9607
and the current Band-Color key legend, here:
http://tradetight.org/forums/thread-view.asp?tid=1445#M9647
... (you can set these up on two separate Browser tabs and flip back and forth easily, btw)
do you think I should use Teal or DarkCyan for the Bull ToeIn Warning color ... OR should I use the existing greenish-Bull-warning color-set?

3. After checking those two snapshots, do you think I should use Magenta or Orchid as the Bear ToeIn Warning color ... OR should I use the existing reddish-Bear-warning color-set?
Top of the page Bottom of the page
KeithBuhl
Posted 6/30/2019 11:55 AM (#9652 - in reply to #9641)
Subject: CVW Journal & DISCUSSION



Veteran

Posts: 128
10025
Location:
USA: PA, New Hope
1. Yes.

2. Teal.

3. Magenta
Top of the page Bottom of the page
RobertPfoff
Posted 6/30/2019 6:02 PM (#9653 - in reply to #9641)
Subject: CVW Journal & DISCUSSION



Friend

Posts: 16
0
Location:
USA: PA, Warrendale
I agree it makes sense to implement the changes

I am already overwhelmed by all of the colors, so why not 2 more. I suspect I will be more reliant on the dots on the charts than all of the color changes but perhaps as I get more comfortable with it I will use it more. It is great to have such a compact way of demonstrating what is going on. Have you thought much of how people who are color blind will deal with it?
Top of the page Bottom of the page
JimDean
Posted 6/30/2019 7:52 PM (#9654 - in reply to #9653)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
Thanks for the reply Keith, and your comment Robert. I am still weighing alternatives here. Tomorrow I will likely post an alternative format which removes five of the colors from the Band - I’ll explain it then.

I did comment in an earlier post about color blindness. Quite a long while ago I had a single client who struggled with that. What I discovered was that so many useful nuances had to be removed to accommodate the limitations, that it just didn’t make pragmatic sense to design “retail”-sales tools around that. I am however open to doing custom alterations per a single clients wishes - but at additional cost.

In this case, the brightness and darkness of the hue should be discernible, even if the red/green distinction is lessened. And generally a glance at the price direction should clarify if it’s bull or bear.

If anyone else has comments or suggestions, please submit them tonight, as I intend to lock this down tomorrow. Thanks!
Top of the page Bottom of the page
KeithParsons
Posted 7/1/2019 10:31 AM (#9655 - in reply to #9650)
Subject: CVW Journal & DISCUSSION



Veteran

Posts: 122
100
Location:
S.Africa: , Umgeni Park, Durban
I would go with Keith
1) yes
2) teal
3) magenta.

The chart dots for me are clear. But then of course and right now I am not under "trading pressure" However and if it was possible it would be most helpful from a visual aspect to have these colors in something like Nirvana's SetUp editor in the F.L.

I do realise that this is not possible.

Regards
Top of the page Bottom of the page
JimDean
Posted 7/1/2019 10:54 AM (#9656 - in reply to #9655)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
Thanks for the reply.

You’re correct. OT does not permit custom coding that leaves the color choices open to arbitrary user assignment. It all has to be hardcoded. And, when many colors interact, meticulously planned.

Regarding clarity - the formatting I’m working on now regarding envelope crossovers will not add the five extra colors described earlier. Instead, I’ve decided to show a set of colored dots immediately under the Indicator Wave Band - a dot appears without incurring or changing that bar’s band color. A dot occurs on bars where the price crosses an envelope on the price pane. I’m in process of considering echoing the existing dots from the price pane to this new “signals row” of dots below the band. Thus, any bar with a dot in that row is one that fires some kind of order.

THEREFORE: every “Signal” has a “dot” somewhere on the Indicator pane OR the price pane (or both), and the multicolor Band clarifies progress as things progress. Also, as mentioned earlier, the user can optionally turn off the drawing on the price pane if so desired.

Top of the page Bottom of the page
JimDean
Posted 7/1/2019 3:00 PM (#9658 - in reply to #9656)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
Knowing When, Why & How to Manually Place Stradicator Orders:

I believe I mentioned during the videos somewhere that the Stradicator paradigm includes the option for popup messages whenever an action(dot) appears on the Last Bar of the plotted Symbol. The message will contain detail similar to the Voteline Advisor pane (+extra) … type and size (%basis) of Order, reason for the action, and a report of prof/loss thus far in the trade (incorporating the dynamic sizing and scaling impacts). The OT voteline Advisor is not able to do this for scaling, btw.

It will also be possible for you to check on what the action (order) and trade status was for some prior Dot. I’m still debating the best way to do this … might be via modifying how the BarWndo slider works, or could be by adding another “BarsB4hre” slider (similar to MTV), or possibly by using a CSV text file to record every event so that it can be readily loaded into Excel. If I cannot figure a way to straightforwardly implement the first of those approaches, I’m more likely to gravitate to the third (Excel file) approach, rather than adding yet another slider. There are pros and cons to all three methods.

A clean and fairly simple First-method solution is explained below, and illustrated in the snapshot.

Currently, the negative-sign for the Master input has limited/doubtful value ... I could reassign the effect of that negative sign to take over what the current BarWndo negative sign does ... that is, toggling on the Equity-Vote Filtering. It actually makes a bit more sense put it at the top, anyways.

Having done that, the negative sign used with the BarWndo input could simply "reverse" the meaning of the input value for the Plot and the Return. When Plot is specified (final input pos), then a negative BarWndo value would "lock in" the number of calc'd-output bars to be 128 (the default 6mo value), but would cause the "Last Bar" calculated ... ie the right boundary of the output window ... to shift to the left ... ie into the past.

This means that the popup message about what trading-action the Dots represent would apply to the final "BarWndo" bar's dot ... to check some particular prior Dot (before the HRE), one would simply move the *negative* BarWndo slider to the left, looking deeper into the past. Each time the right edge of the window finds a bar with a Dot on it, the Popup would appear. To move on, the user clicks OK on the popup, and adjusts the slider once again.

This offers a convenient way (fewer intermediate "OK" clicks) when a particular bar is in question that's several "Dots" ago: with the Indicator Param panel open, the user first changes the Plot output slider to a non-Popup mode, then move the neg BarWndo slider until its right edge is on the desired Dot, then move the Plot output slider back to Popup mode to learn the details. This could be done repeatedly. When finished, clicking "Cancel" for the Indicator Param Pane would restore the output to its original configuration. Not *quite* as simple as clicking on a Voteline triangle to view the Advisor window, but with two huge advantages: 1) no need to recalc the strategy since Stradicator is self-contained 2) Strad Popup provides info that the Voteline Advisor simply *cannot*.

I'd value your *immediate* feedback on this idea, if you have comments. Thanks.



(Neg-Master+BarWndo Changes.png)



Attachments
Attachments Neg-Master+BarWndo Changes.png (147KB - 0 downloads)
Top of the page Bottom of the page
RobertPfoff
Posted 7/1/2019 9:20 PM (#9659 - in reply to #9641)
Subject: CVW Journal & DISCUSSION



Friend

Posts: 16
0
Location:
USA: PA, Warrendale
I think this is an excellent solution. I have found the dots to be very intuitive and the addition of this level of information, especially when paired with an equity curve will be amazing. Keep up the great work.
Top of the page Bottom of the page
JimDean
Posted 8/17/2019 8:53 PM (#9670 - in reply to #9658)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
Howdy all ... finally getting back into things, after a long and slow improvement in a health issue ... and a few family issues. Hopefully, as things get rolling again, y'all will be able to respond in a timely fashion to questions posted here for your input (ie within 24 hours or less) - thanks. If I don't get timely responses, then I'll spend less time posting things. Up to you as to how much time you'll be able to spend ... any you can will be very helpful.

Today, I got the changes in my prior post implemented:

Negative Master input now acts as a flag to Engage the Equity-Filter-toggling
... note that Equity-Filter toggling is an optionally-licensed feature for *any* version
... Summary of version diff's, here: http://tradetight.org/forums/thread-view.asp?tid=1445#M9669

Negative BarWndo input now serves to "flip" the Plot vs Return application of the input value:
... Pos BarWndo for plot = width of eval, thru HRE; Neg BW for plot = #BarsB4Hre, w/128 width
... Pos BarWndo for rtn= #BarsB4Hre, w/9-bar width; Neg BW for rtn = width of eval, thru HRE

And, I formalized some licensing flags regarding the use of the optional Equity Filter toggles as well as the (newly) optional Stradicator-Filtering specs (via the two Guru inputs). Either or both of these features can be used with any of the Versions, but they'll be independently licensed for an extra fee (since they are not only powerful and unique, but also pretty tricky to code).
Top of the page Bottom of the page
JimDean
Posted 8/17/2019 11:35 PM (#9671 - in reply to #9670)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
Later addendum to prior post:

If you read my prior posts carefully, you’ll note that there is another important reason for modifying the meaning of the negative-input flag for the BarWndo parameter … when the flag is negative AND a Plot Output is specified (final input>0), then that toggles ON the Popup Advisor feature - if the final output bar has any type of Signal, the Advisor explains what to do to implement the signal, in full detail (buy, sell, size, etc) - also, it explains the cause for the signal and recaps the trade progress thus far. The negative BarWndo value controls which bar is in focus (to examine the signals for the fifth bar before the HRE, input -5).

I also previously discussed the possibility of creating a log file output in CSV (Excel) format, that would essentially mirror all the Advisor info for every calc’d Signal for every symbol in the FL, for the number of bars specified. Obviously this could be very useful, but due to the potentially large amount of file-writing, it might slow things down. So, the jury is out on that one. If I do implement it, I’ll create a new Final-parameter input option to toggle the Log output - maybe via a Help panel button. Not sure yet.
Top of the page Bottom of the page
JimDean
Posted 8/18/2019 8:45 PM (#9672 - in reply to #9671)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
This post requests feedback ... please respond ASAP.
If you don't understand the Q, then please ask for clarification.

READ THE UPDATED DESCRIPTION OF FIVE OPTIONAL-LICENSE FEATURES, HERE:
http://tradetight.org/forums/thread-view.asp?tid=1445#M9707

=============

QUESTIONS:

1. Are those descriptions adequate ... do you "get it" what the flag is toggling, and do you at least roughly perceive that each of them offers a lot of potential benefit?

2. Do you think that some users might want to buy the Strad without some of those features, in order to save some money? (no idea yet but each feature might cost $200-$300 or so).

3. Any other comments are welcome. Thanks!

==============

Attached snapshot shows how the first Help panel will indicate on the second line, which features are active for that user ...

(Help Panel #1 - ExtraF.png)



Attachments
Attachments Help Panel #1 - ExtraF.png (416KB - 0 downloads)
Top of the page Bottom of the page
JimDean
Posted 8/19/2019 3:08 AM (#9674 - in reply to #9641)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
I suspect that every one of you who bought in early are interested in most all of the features.

The point of my three questions was to survey y’all re general understanding of the five descriptions (ie enough to make you want to know more or not).

Also, it would *really help* if you’d tell me what you think the broad swath of *all* future clients might want (or not want). Your best guess.

Reason: if virtually all clients will want virtually all the options, then I don’t need to offer a buffet. But if the all vs none price has a difference of say $1000 per Stradicator, do you think some clients might prefer to pick and choose some and not all?

A client will be able to try out the full-featured engine out before committing with significant money. Prospective buyers can “rent” the full-featured Stradicator for a month (or six months), at a small fraction of the cost - then once they’ve decided what features they find useful, they can purchase that particular combo. It’s easy to add more later.
Top of the page Bottom of the page
RobertPfoff
Posted 8/19/2019 10:39 AM (#9675 - in reply to #9641)
Subject: CVW Journal & DISCUSSION



Friend

Posts: 16
0
Location:
USA: PA, Warrendale
I think combining the features makes the most sense, it is hard to grasp how useful each feature is until you use it. The one that makes the most sense to break out is the Wizard-Config.

The holy grail may being configuring each of the stradicators as super inputs into the GA. It sounded like that could be done. Is that your understanding?
Top of the page Bottom of the page
JimDean
Posted 8/19/2019 11:15 AM (#9676 - in reply to #9675)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
The Strad’s can be used in a GA, but not in SETI or anything that requires uploading to N’s servers. N prohibits DLL uploads, sad to say.

However, the potentially most beneficial and profitable features of Strad’s, which are the equity-filtering and the equity-stats-based dynamic OScans, would not be applicable to GA’s.

The whole GA approach uses a nearly 180-degrees opposite methodology than Stradicators. GA is essentially a supercharged Avengers-level curve fit. Stradicators are based on cold logic, therefore can be expected to be more consistent and robust. (Jmho).

The “N method” is primarily to build cool tools that work with a static (or simplistically dynamic) focus list. The Stradicator method is to work with an extremely-insightful dynamic-Scan Focus list - that is, find the symbols that historically worked best with the rule, rather than the other way around. And redo that every day/week.
Top of the page Bottom of the page
RobertPfoff
Posted 8/19/2019 2:32 PM (#9677 - in reply to #9641)
Subject: CVW Journal & DISCUSSION



Friend

Posts: 16
0
Location:
USA: PA, Warrendale
I agree that the Strad's are significantly different from the GA, that is why I like them so much. That is also why I think they may be good inputs into the GA. I understand why they can not be uploaded to the Nirvana servers for SETI, but it seems to me that they should still be callable as individual inputs for the GA but I don't understand the inner details of how that all works. Thanks
Top of the page Bottom of the page
JimDean
Posted 8/19/2019 3:13 PM (#9678 - in reply to #9677)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
Hi Rob - once they are released I’ll explain how. It’s easy - just create a simple OScript formula (Boolean or Value), which includes a function call to the Strad indicator with the appropriate return value selected. Many variants possible. You could set it up to be GA’ing for entry signals, or as a Filter.
Top of the page Bottom of the page
DonTurner
Posted 8/19/2019 7:31 PM (#9679 - in reply to #9641)
Subject: CVW Journal & DISCUSSION



Spectator

Posts: 3
0
Location:
USA: CA, Ridgecrest
Jim Dean said, "Also, it would *really help* if you’d tell me what you think the broad swath of *all* future clients might want (or not want). Your best guess."

My best guess: The probability distribution of quantity of features for future clients might be similar to the distribution of $ spent by us early adopters. I especially like the equity influenced and dynamic focus list features.
Top of the page Bottom of the page
JimDean
Posted 8/25/2019 7:25 AM (#9683 - in reply to #9641)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
Steve’s likely too busy to get into this - but I agree that we need better analytics. To that end, with a focus on how much useful info I can derive from a single Stradicator run on a single Symbol (especially for fitness Scans), I’ve come up with a basket of metrics that I feel will be far more representative and effective than the classic APR, PPT, MDD, Calmar, etc. These stats will be part of the Paradigm, available as Return value output choices, for Scans, Focus List columns, etc:
http://tradetight.org/forums/thread-view.asp?tid=1445#M9645

I’ll be asking Mark to fold them into the spreadsheet as well, modified appropriately to encompass multiple symbols. I’d also like to incorporate some frequency distribution and error-estimate info to help qualify them further.
Top of the page Bottom of the page
JimDean
Posted 8/25/2019 8:01 PM (#9688 - in reply to #9683)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
I just moved the last several posts that have to do with Log Files and StradFolio into a new breakout thread, here:
http://tradetight.org/forums/thread-view.asp?tid=1446

Please post about that topic in the new thread. Thanks.
Top of the page Bottom of the page
JimDean
Posted 9/9/2019 6:50 AM (#9706 - in reply to #9688)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
I just updated some of the crib sheet info and panel snapshots in the Reference thread, to reflect recent changes in how the BarWndo input works (pos vs neg, and override), including the shift of the EquityFilter toggle to the Master input. Note that the OmniScript function-call panel now has expanded detail to match. These posts were affected:

http://tradetight.org/forums/thread-view.asp?tid=1445#M9633

http://tradetight.org/forums/thread-view.asp?tid=1445#M9634

http://tradetight.org/forums/thread-view.asp?tid=1445#M9635

http://tradetight.org/forums/thread-view.asp?tid=1445#M9642
Top of the page Bottom of the page
JimDean
Posted 9/9/2019 3:24 PM (#9709 - in reply to #9706)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
Revisiting the Strad-Paradigm Parameters ...

Currently, two of the fifteen param-panel rows are used for "dummy" parameters that are greyed out and serve only as separators. Their value is primarily aesthetic.

I'm thinking about stealing one of them, so that the equity-filtering feature can have a slider to allow the user to adjust it from conservative to aggressive in some way. The equity-filter watches all the results of trades for a given symbol over time (for those particular strad input-settings), and decides when the performance has for some reason become less than acceptable. At that point, it keeps on calculating all the trades "in the background", but until the performance gets better, it blocks any of those background trades from being actually processed as orders (ie outputs).

The equity-filter logic as I previously implemented uses a stack+slope voting algorithm, using the actual (background) equity curve with two MA's and some simplified rules. There are several internal variables that drive the calc, such as MA periods, Slope periods, vote thresholds, etc. The slider that I'm referring to would allow the user a "guru" type input which would adjust all those params in an appropriate manner so that the equity filter would be more or less reactive to performance changes, before it blocks the trades.

Currently, the *only* way for the user to tweak that *requires* that they own the Cstm version, and that they use a manual-input Override of the final Output (Return|Help|Plot) parameter to adjust things. I've never liked this much, for a whole lot of reasons. Here is what the current Override spec for that input looks like:
OutputReturnHelpPlot- EqtyVote IRCVS##: ##=PltRtn; S~TimeG, V~VoteG, C=Cntl
... I&R=SmartSize Invst&Risk Ntry +/-10% from ATR/5% study (RobustTP only)

Since the +/- and two of the digits must be reserved for the actual Output selection, that only leaves five to control the other specs ... and it's a pain to change them ... and when in use, that in turn makes it a pain to adjust the Output (diff Return values, diff Plot formats).

SO ... if I create a new slider dedicated to Equity Filtering (which needs to be licensed to work), it would leave the final Output slider alone for convenient output-switching, and it would provide a friendly guru-control to adjust the filtering rather than locking in on one arbitrary set of rules. This could be made available for all Versions if that makes sense to do ... and the Override would be avail only for the Cstm version ... and it would have all seven digits as well as the +/- to work with.

Seems to me to be a no brainer ... losing an aesthetic separator to improve functionality of one of the major Stradicator features.

Do you think I should steal a divider-line to make this Equity-Filter Control slider?

------------------------

And, while I'm at it ... there's one other issue that could be handled effectively by creating yet another slider ... stealing the other aesthetic separator ... that being, an input of the Bar Length (ie 1 day, 20 min, 4 weeks, etc). This input would not actually force OT to use a different timeframe ... you have to do that with the DataPeriods table and the dropdown menu (or with the custom Strategy timeframe spec). However, there are a lot of other benefits:

1. It will significantly speed up program execution when more than one barlength is defined in DataPeriods. Currently, even though there's no need to, OT insists on running each FL symbol for every defined barlength, even if that barlength is NOT the "current" one which the chart is showing. But if the Strad has an input saying "THIS run is targeting a PARTICULAR barlength", then I can easily add code to completely skip the Strad calc for all other barlengths. This can be a major time-saver.

2. It's not at all unreasonable to presume that you might want to use a different pattern of parameter settings for different barlengths, with the same Stradicator. There could be all kinds of reasons for this. By having an input which specifies the intended Barlength, that prevents you from accidentally using that input-pattern with other barlengths that wouldn't be a good fit.

3. If the BarLength is part of the Strad's input-param list, then the "Pattern-ID" spec that is central to the TradeLog file management will contain ALL the relevant distinctions for that management. This simplifies file-naming, etc somewhat. Not a good enough reason all by itself, but a beneficial by-product.

This input, a slider from -20 to +20, will be available in all Versions. Input=0 means accept all barlengths.

Negative slider values will map to Minute-barlen's. Barlen's which divide evenly into the 390 min of a trading day= 1,2,3,5,6,10,13,15,30,65,130,195 ... round-numbers that someone might choose= 20,40,45,60,75,90,120,150. A total of 20 options. These will be mapped: -1 to -20 = 1, 2, 3, 5, 6, 10, 13, 15, 20, 30, 40, 45, 60, 65, 75, 90, 120, 130, 150, 195.

Positive slider values would be for Daily-or-higher barlen's ... including multiples of daily|weekly and fixed monthly (no mo-multiple avail). Logical limits for each will define the mapping:
1-9= #daily bars, 10-13=1-4wks, 14=1mo, 15=Volume, 16=Range, 17=3-LineBrk, 18=Renko, 19=Point&Figure, and 20=Heiken-Ashi, to correspond to OT's Periodicity -1>-3 and -5>-7.

The advantage to this mapping is that it encourages users to pick "rational" barlengths. The mapping will in a Help Popup table, and the first Help panel will indicate the actual #minutes.

Do you think I should steal a divider-line param to make this BarLen slider?
Top of the page Bottom of the page
KeithBuhl
Posted 9/9/2019 10:14 PM (#9710 - in reply to #9641)
Subject: CVW Journal & DISCUSSION



Veteran

Posts: 128
10025
Location:
USA: PA, New Hope
Equity Filter Control Slider ---- YES

BarLen Slider ----YES

Cheers,
Keith

Edited by KeithBuhl 9/9/2019 10:16 PM
Top of the page Bottom of the page
LDNewby
Posted 9/10/2019 2:52 PM (#9714 - in reply to #9709)
Subject: CVW Journal & DISCUSSION



Friend

Posts: 36
25
Location:
USA: VA, Vienna
I concur with Keith

Yes to the Equity-Filter Control Slider

Yes to Bar Length Slider.
Top of the page Bottom of the page
MarkHolstius
Posted 9/10/2019 3:09 PM (#9715 - in reply to #9709)
Subject: CVW Journal & DISCUSSION



Veteran

Posts: 206
100100
Location:
USA: IL, Sleepy Hollow
Yes to both...
Top of the page Bottom of the page
RobertPfoff
Posted 9/10/2019 9:59 PM (#9717 - in reply to #9641)
Subject: CVW Journal & DISCUSSION



Friend

Posts: 16
0
Location:
USA: PA, Warrendale
They both sound useful to me, If I had to choose I would favor the equity slider over the bar length option
Top of the page Bottom of the page
KeithParsons
Posted 9/11/2019 1:57 AM (#9718 - in reply to #9641)
Subject: CVW Journal & DISCUSSION



Veteran

Posts: 122
100
Location:
S.Africa: , Umgeni Park, Durban
Both please
Top of the page Bottom of the page
JimDean
Posted 10/2/2019 2:40 PM (#9731 - in reply to #9709)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
In light of the comments in this post:
http://tradetight.org/forums/thread-view.asp?tid=1446#M9729
... it appears that the reasons for the barlength slider (mentioned in my prior post, above) have gone away.

The rationale for the new Equity slider remains valid ... and this post:
http://tradetight.org/forums/thread-view.asp?tid=1446#M9730
... explains how the Master-input negative-sign could be quite valuable for runs using independently-tuned LongOnly and ShortOnly input-patterns.

So, the existing two gray "separator" inputs will go away. There will still be room for one more parameter in the single-column pane ... it makes sense (to me at least) to use it to visually separate the inputs that are "specific to" the particular Stradicator (CVW, ADX, etc), from the inputs that are part of the "paradigm" ... ie essentially the same meaning for all Stradicators.

In the case of CVW, that gray separation-param would be just before the VoteGuru input ... all the inputs above that are specific in meaning to CVW. (the new Equity slider would be below it, in the Paradigm section)

ALSO ... the fifth "optional features flag" mentioned here:
http://tradetight.org/forums/thread-view.asp?tid=1444#M9672
... instead of toggling "wizard mode", would activate the ability to run the dual-direction custom-models described above.

Comments welcome.
Top of the page Bottom of the page
JimDean
Posted 10/6/2019 10:43 AM (#9743 - in reply to #9731)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
This is an IMPORTANT post, for those of you who are trying to follow along ...

In a (big) nutshell, the changes mentioned in the prior post, and documented in the link it provided, were occasioned by these considerations:

1. A slider to control Equity Filtering impacts really was needed, since it's a major Stradicator Paradigm feature, and has significant impacts ... and that slider provides for manual override inputs in a logical place, rather than cumbering the OutputReturnHelpPlot slider with them, which should make it *easy* to flip between alternative plot formats, or open Help, sans manual typing.

2. TradeLog output opens up an exciting opportunity for StradFolio and StradWiz spreadsheet evaluations ... both of which can benefit from flexible control of the historical periods being backtested, and flexible spec of trade-directions allowed. The existing slider did a lot, but seemed constrained ... splitting it into two (like MTV) seemed a way to fix that ... and it provided extra "room" for CutGrab manual-overrides, albeit in a non-intuitive location. (see new idea, below).

3. The idea of allowing separate Long vs Short-Trade tuning for the four Paradigm Expert-inputs (Bgn/End/Fuz Vote & CutGrab-envelope), either via neg Master or slider-mod's, seems to have a lot of merit (tho, sadly, little team-feedback ... my bias is to make the variants transparent rather than using the neg Master approach). This is comparable to setting up an OT Strategy diagram with two "rows" each dedicated to a separate trade direction.

4. Long vs Short tuning logically should "scale" like other version-features ... the Cstm version logically should allow manual-overrides for relevant impacts. My initial attempt (see link above) to tailor that by adding "lower-case" digits to the B/E/F slider override-spec's worked out OK, *but* the complex nature of what the CutGrab slider does made it impossible to handle Long/Short override distinctions on that input alone ... I had to split out some of it to the WndoWdth and BarsB4hre sliders. Very non-intuitive ... but it was getting late. (see new idea, below).

======= improved approach after sleeping on it =======

A. BarWndo slider ... I was too hasty in expanding this to two sliders. It still may be feasible to provide both "easy-slider mode" for output ranges in "normal" situations, AND also provide sufficiently-flexible "TradeLog backtest specs" via manual-typed spec's.
Part of the issue is that a six-digit override "GggWww" (for #Width-bars and #Gap-bars) is limited to ~4 years of output-window width, &/or ~4 years historical lookback "shift". OT has daily-bar history that goes back to 2000 ... approx 20 years (5000 bars) ... and has enough intraday history to provide for a few tens of thousands of bars (N keeps changing the max ... not sure what it is right now but probably at least 20,000).
I tried to deal with that by providing four Gggg digits (push the window further back in time) ... but that still precludes a daily-bar output-window width (Www) of more than four years ... OT inputs are limited to 7-digits.

This morning I thought of a simple work-around for the backtest-window GggWww limitation ... allow an optional seventh "Multiplier" digit (MGggWww): M=0 (or 1) means Ggg & Www represent a normal # of bars ... input of "123456" (or 1123456) says to model an output window that is 456 bars wide, that ends 123 bars before the HRE.
For M > 1, the Ggg and Www spec's deal with more than single-bar increments. When M=2, a MGgg=2123 input specs a 246-bar Gap. MGgg=4321 spec's a 4*321=1284-bar Gap, and so on. Thus the max "reach" of MGggWww, using a manual override of 9999999 specs a window approximately 9000 bars wide, ending 9000 bars ago (9000 daily bars = 36 years; 9000 5-min bars = nearly six months). That provides PLENTY of range ... and it's easy to see where the "window" lies, simply by applying the MGggWww input to plot mode (with adequate DataPeriods loaded). Yes ... when M>1 there is some loss of "precision", but backtesting isn't *supposed* to be "precise".

Another important backtesting trick, especially when trying to discover what input-Pattern works for particular market conditions (ie bull or bear), is to visually search out multiple "chunks of time" when the market is behaving in the desired fashion. Determine that appropriate MGggWww spec to encompass each "chunk". Create a TradeLog file by running your Strad and Symbols for that time-chunk. Repeat this for each different MGggWww "chunk". Then, you can easily merge (concatenate) those TradeLog files together into a common StradFolio or StradWiz analysis spreadsheet. Voila!

So ... the MGggWww spec removes the broader-backtest-range need for separate sliders.

The other thing that single-slider controls is allowable Trade-Directions. Default is "Both" ... ie Long & Short trades allowed. The slider itself goes to 256 (ie one year, which is a reasonable max to view on a plot at one time ... or a reasonable max lookback to use for a FocusList-Column Return value.
The "BarWndoTrdDirL3cS6c" parameter-name reminds the user that when the (3-digit) input is 300-599, the model is LongOnly; when it's 600-899, the model is ShortOnly. The param-name hopefully makes this fairly easy to remember.
Negative 3-digit inputs simply reversed the meaning of the value based on the specified Output type: for plot, positive=width@HRE and neg=lookback(W=128); for return, pos=lookback(W=9) and neg=width@HRE. This was explained at length when I came up with it ... a clean and elegant solution that handles 95% of the uses people are likely to need.

When a manual override MGggWww is used, the earlier definition limited the model to LongOnly for positive inputs and ShortOnly for negative inputs. In retrospect, that was a poor choice, since there are probably no (successful) traders that limit themselves just to Short trades.
It makes more sense for a positive value to represent Both Long&Short, and a negative value to limit the trades to Longs Only.

But ... how can the new "separate-pattern" Long & Short models be specified, distinctively from the "common-pattern" BothLS default-approach described above?

Yesterday, the need for that distinction was part of what pushed me to split the BarWndo input into two separate sliders ... but that split, per notes above, is no longer in play. It turns out that the solution is SIMPLE, even using the single-slider ...
When the BarWndo slider (or manual override) implies "BothLS", then use a *common* pattern if none of other Strad inputs imply separate L vs S models ... but use *distinct* L vs S patterns if any of the other inputs provide those distinctions.

The "distinct L vs S input spec" could be via the neg-Master approach (if that's implemented ... looking like it won't be) ... or could be via "integer.decimal" inputs for the four Vote+Envelope Experts, as described here ... http://tradetight.org/forums/thread-view.asp?tid=1446#M9739
If any of those V+E inputs tell the Strad to evaluate Long vs Short differently, and if the BarWndo slider indicates "BothLS" modelling, then L & S trades have different entry and management rules. If none of those inputs imply separate evaluations, then L & S trades use common rules.

This solution also provides a "backdoor" way to do Shorts-Only TradeLog runs in conjunction with a BarWndo MGggWww override. Under the revised rule above, when that override is negative, it normally spec's a LongOnly model. However, if any/all of the "integer.decimal" V+E Expert inputs have nonzero decimal (Short Trade spec) and zero integer (Long Trade spec) digits, then it's logically clear that the intent is to run Shorts-Only instead of Longs-Only.
This trick is likely *only* needed when doing research to find an "optimal" input-pattern for Short trading ... in actual practice, the Strad would be set up for BothLS with separate eval's, and both integer and decimal digits would be nonzero. I realize this trick is sort of complex ... but hey, it's definitely one that only power-users might need. I'll create a how-to video explaining the whole thing, fear not.

A. Summary ... all that discussion results in the conclusion that only *one* slider is needed for BarWndo control ... it will use the same 3-digit pos & neg input rules as before ... and the manual override will implement the "M" digit for expanded range, with +override meaning BothLS, and -override meaning LongOnly (unless integer.decimal Experts have integer=0).

======= further next-morning thoughts re Expert overrides =======

B. CutGrab overrides ... as explained in #4 above, when I tried to create separate Long/Short manual-overrides for the CutGrab slider, there weren't enough digits available for that complex spec ... so I had to "rube-goldberg" it by pushing four of the necessary digits into totally unrelated sliders (the two that controlled BarWndo). However, per A above, I've decided to go back to just one BarWndo slider ... which has absolutely no extra override digits available to "help out" the CutGrab spec.
The original CutGrab-envelopes override-spec presumed the same pattern is used for both L & S trades ... it used all 7 digits to provide full user-tailoring of the powerful rules which control those envelopes ... here is the full existing spec for that input:
==pos== CutGrabGapSzSm2Lrg- pos FullX: 1-5 => incrs Gap, decrs Tytn (0=off, 6-10=CutOnly)
... InitGap ATRmult 1-5: GrabProfFX= 4x,5x,6x,7x,8x; CutLossFX= 3.2x,4x,4.8x,5.6x,6.4x
... FastTytn%InitGap/bar: 1-5= -24>-20%; Slow=half if abs(G-G1)<WTR/2; 0.x=> AllOff
==neg== CutGrabGapSzSm2Lrg- neg 1-5 same FullX C+G, plus PXit Cut+Grab (neg 6-10=CutOnly)
... PC&PG def=PXsize, gap=FX/2; PC=dkGrayPX, PG=ltGrayPX(notWarn,noAdd); FC&FG=blackFX
==ovrid== CutGrabGapSzSm2Lrg- GPFCAUT: init gap ratchets w/price & tightens with time
... T=FstTytn/bar 0-9= 0|18>26% *InitGap; G=GrbS & C=CutS XitSiz 0-5= PX|10-50%;
... Part+Full 0-9 (0=off) ATRgaps*: Grabs P= 1-4x, F=2-8x; Cuts A= .8-4x, U= 1.6-8x
... Use EndV's PXsize for G or C =0; set InitGapMult=0 to deactivate specific line(s)


Note also that even the "standard" (non-override) spec for this slider had some extra tweaks ... input-values of 1-5 specify Cut & Grab envelopes, while (parallel) 6-10 input-values limit it to just Cut envelopes, but use gap-spacing similar to 1-5. Negative slider values add Partial-Exit envelopes to the Full-Exit envelopes that pos or neg slider values define. Yes ... this is a lot to expect the user to remember ... fortunately, as the slider is adjusted, the lines appear and move around on the plot in an obvious fashion ... which makes things more intuitive.

The combo of the pos & neg slider functionality, with the powerful-but-complex 7-digit manual override makes it pretty clear, in retrospect, that the CutGrab slider really should be TWO separate sliders ... one controlling LossCuts, and the other controlling ProfitGrabs. Since there's now room for one new slider ... this is the way to go. It simplifies and clarifies the entire issue.

Also, to further simplify use, I'm bringing the CutGrab controls under the wing of the VoteGuru. That is, when either of the CutGrab sliders are set to zero, the 0-input's spec will be dictated in a logical fashion by the value of the VoteGuru "Tyt2Lax" slider (just like the Bgn/End/Fuz sliders are, now). This keeps the "Easy" version clean ... no separate CutGrab sliders will show up there (Xprt, Powr, and Cstm versions *will* show CutGrab sliders). When VoteGuru is "Tight", the envelopes are too ... when it's "Loose", so are the envelopes. Simple, and logical.

With this mapping to the VoteGuru, and also with the integer.decimal L.S rule in play, it means that the LossCuts and CutGrabs sliders will use a range of -6 to +6. Zero means "use the VoteGuru rule". Positive specifies Full-Exit envelopes; negative spec's both Partial & Full-Exit envelopes. 1 > 6 adjusts the envelopes from "Tyt2Lax" (like the VoteGuru), gradually moving them farther away from the candles. +6 turns off the adjustable Full Exit line; -6 turns off Full Exit line but keeps the Partial Exit in play; +/-6 for Cuts retain a wide-gap fixed Broker-Stop line; VoteGuru doesn't ever select +/-6 automatically ... that requires nonzero Expert input for LossCuts &/or ProfitGrabs sliders.
The +6 > -6 range is all single-digit ... therefore, the previously-discussed "integer.decimal" shortcut for distinct L vs S modelling will work fine here.

The many internal factors that affect the formula which are currently keyed to a 1-5 range, as documented above, will remain as is. But since there are two sliders now, that "makes room" for "clean" explicit Long vs Short manual overrides for all those factors.
The existing 7-digit override is GPFCAUT ... GPF applies to Grabs, CAU applies to Cuts, and T is applies to both. G/C define Partial-Exit size; P/A define Partial-Exit gap-distance; F/U define Full-Exit gap-distance; T specifies how quickly the envelope-gaps "tighten" over time. However, there is no way to indicate how these spec's should differ for L vs S trades ... and it certainly does make sense that the envelope-gaps &/or PXit-sizes &/or tightening might optimally differ for L vs S.

The new separate LossCuts and ProfitGrabs sliders can each have their own 7-digit override. It seems clear that the Full and Partial Gap-distances should be separately spec'd for L vs S (2*2=4 digits). Tightening currently is T=FstTytn/bar 0-9= 0|18>26%. PXitSize currently is 0-5= PX|10-50% (0=noPXits). The 7-digit OT limitation means that one of these can be independently L vs S controlled, but the other will function the same way for L & S.
Generally, S trades "move faster" than L trades (exceptions found in an extended bearish markets) ... there are usually fewer total bars in a S trade. This implies that there will be fewer opportunities for PXits ... arguing that they may need to be bigger for S than for L trades.
OTOH, Tightening is a relatively "fine" and somewhat arbitrary adjustment (tho considerable research went into it) ... the median value is a reduction of 24%/bar when "fast" conditions warrant (too complex to describe here). The definition of "fast" already adapts to the typical "speed differences" of L vs S moves ... therefore, a separate L vs S spec for T would be redundant.

The 7-digit override for each of the envelope sliders will be spfSPFT, where T is tightening for all cases, SPF is for Longs and spf is for shorts. S|s defines the size of partial exits; P|p defines the gap for partial exits; F|f defines the gap for full exits.
Hopefully that extended explanation is clear, and has helped you better appreciate the smarts behind these features, as well as the degree of control available to the user.

B. Summary ... the CutGrab -10 > +10 slider becomes twins: LossCut and ProfitGrab, each -6 to +6 ... they are both brought under the aegis of the VoteGuru slider mapping, when their inputs are zero, and neither shows up in the Easy version. Cstm version manual overrides are easier to understand and allow fully-distinct spec of how envelopes work for Long vs Short. Integer.Decimal inputs are allowed as well, for separate L vs S expert models.

====================================

WHEW! I think that finally closes the revision-loop which started off due to the separate Long vs Short idea, and the TradeLog backtesting-tool opportunities. The net result is, I believe, easier for the user to understand than before (Equity slider separate, and Cut/Grab independent but under Guru control) ... with far greater modelling flexibility.

------------------------------------------------------------------------

Based on comments above, and extensive discussion in the TradeLogs thread, the Positive, Negative, and Override sections of the Input-Parameter Definitions have been "updated" here:

http://tradetight.org/forums/thread-view.asp?tid=1445#M9625

Comments and questions welcome ... but please read this at least twice, first ;~)
Top of the page Bottom of the page
JimDean
Posted 10/8/2019 8:20 AM (#9745 - in reply to #9743)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
Almost finished with the TradeLog engine ... here are some reminders re its usefulness:

1. TLB output file provides full history of all bars of all Symbols during all Trades that use a specified Strad input-Pattern across a specified test Window for a given bar-length

2. TLS output file provides 13 metrics, both "classic" ones and TradeTight-exclusives, derived from all the trades in the parallel TLB file, listed separately for each Symbol

3. TLB output is designed to "feed" the StradFolio analysis spreadsheet ... a major expansion of Portfolio Simulator to handle the Stradicator-Paradigm's special capabilities

4. TLS output is designed to "feed" the StradWiz analysis spreadsheet ... a Stradicator-Paradigm expansion of Strategy Wizard, used to find "optimal" modelling-patterns

5. Both TLB & TLS files are in standard "csv" (comma-separated-variable) Excel-import format ... users can easily design custom spreadsheets of their own to utilize them

6. Check out the TLS file's list of an organized collection of statistical metrics which are used by StradWiz ... updated today ...
http://tradetight.org/forums/thread-view.asp?tid=1445#M9645

(many thanks to Mark Holstius for his commitment to help develop the StradFolio sheet)
Top of the page Bottom of the page
KenWilsdon
Posted 10/9/2019 8:57 AM (#9746 - in reply to #9641)
Subject: CVW Journal & DISCUSSION



Friend

Posts: 47
25
Location:
: ,
Looking at the "can be calc'd " values only, if I want a quick look at the values and for some reason do not want nor have time to run StradWiz because of other engagements, I would like to see in the file:

NetPL = SPL (can be calc'd)
ProfFac = PF (can be calc'd)
APLvsADD = APVD (can be calc'd)
Calmar = Calmar (can be calc'd)

The others could go to StradWiz as far as I am concerned (afaiac).

Thanks for the feedback. That kind of info is helpful to me re what the Return Value options should be for the Easy/Xprt versions of the Strad ... ie what the user might choose for output to the Focus List, or for use in an OmniScan.
OTOH, the TradeLog output files ... specifically TLS in this case since it provides all the stat's from each Symbol's run ... are THE SOURCE for any StradWiz analysis. As such, if a user might want to see any value in StradWiz, then that value must be present in the TLS file ... or it must be easily derive-able from other values in the StradWiz file. For instance, if StradWiz includes NetPL and also SumProfTrds, it's easy to calc SumLossTrds = NetPL-SumProfTrds.
Top of the page Bottom of the page
JimDean
Posted 10/9/2019 11:27 AM (#9747 - in reply to #9746)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
Ken ... note reply in blue to your post above ... I'm trying to keep responses together in context.

I've been giving a hard look again to the Metric Stats that will be calc'd by the Stradicator, and made available individually via Return options for FocusList, OmniScan, etc function calls ... as well as en masse via the TradeLog TLS output file. I did a major revision to the post that is linked from several places ... the original content is still there but it's small and hard to read on purpose.

I hope that the revisions are a bit easier to follow.

Any of final six values are, I believe/hope, very good overall measures of the performance of a given Stradicator pattern-model, for a given Symbol, across a given test-window. Time will tell which of those I recommend most strongly. I'm not sure yet which one(s) of them will be avail to the Easy version ... the Cstm version will offer them all ... Xprt & Powr will offer intermediate subsets.

http://tradetight.org/forums/thread-view.asp?tid=1445&#M9645

Note re CAGR & Calmar:
For reasons that require a long explanation ... having to do with the "evils" of Compounding for backtest analysis, and the "pitfall" of metrics that vary based on initial Account Size, and the "unrepresentative" nature of MDD & annualization ... I might decide to *eliminate* CAGR & Calmar completely. If I do include them, I will need to *modify* the CAGR formula somewhat to fit into the Stradicator Paradigm's %Basis "sizing". I've emailed a couple of you about this in detail, for comments and suggestions.

Note re "Holmark" metric:
Mark Holstius did a huge number of tests to develop an empirical formula that I'm calling "Holmark" (sounds like Calmar and hallmark, as well as Mark's name ;~). The tests were based on tens of thousands of RTM trades over many symbols and years. He restricted the terms of the formula to components that are available from PortSim outputs (HitRate, SumWins%, SumLoss%, #Trades, SumPL%, ProfPerTrade, and AvgLoss%).
I'm not sure yet if I'll include Holmark in the Return-Value options and standard TLS file output. It definitely will be part of the StradWiz and StradFolio sheets. For now at least, I'm not publishing Mark's formula. It's an excellent empirical fit ... with a huge database ... but since it's based on RTM strategies that have short hold-times and their own particular "take" on things, I'm not sure if it will work as well with other types of trades ... the confidence we have is based on visual evaluations, rather than theory. By contrast, the formulae in the linked-post above have not been tested *at all* (yet) ... but they have a clear and logical theoretical basis.
Thanks, Mark, for offering this!

Top of the page Bottom of the page
JimDean
Posted 10/12/2019 12:49 AM (#9748 - in reply to #9747)
Subject: CVW Journal & DISCUSSION



Owner/Admin

Posts: 3856
200010005001001001002525
Location:
USA: GA, Lawrenceville
Boy am I glad that Ken Wilsdon is back from overseas ... he is a fantastic helper, especially when I get mired down in decision-making about complex details. All of you should thank him for devoting countless hours of phone time to assist with this project!

After repeated rethinking and extended conversation with Ken, I finally feel really good about the collection of metrics that the Stradicators will calculate, and make available via Return values as well as in the TLB (StradFolio) and TLS (StradWiz) files. I've updated the reference list once again ... here is the link. This is important and to some degree groundbreaking stuff, so it would be great if many of you could chew it over and ask questions or make comments ...

http://tradetight.org/forums/thread-view.asp?tid=1445&#M9645

Here are some notes about the recent decisions ...

1. For each Trade, I'm already reporting NetPL as well as WmaTR & TradeQuality (fancy denominator formula). I'm going to also report the TrdMxDD for each trade ... together, these allow simple external calc (from TLB file) of NetPL/WmaTR or NetPL/TrdMxDD if desired (TrdMxDD has a "Band" in #2)

2. I decided not to report StdDev or Variance or WiggleAvg ... all are messier to calc than I think their value is worth, especially since the TLS metrics are Symbol-specific, and in many cases the sample-population will be very small. Instead, I'm providing a "Band" value in addition to the related Average ... the Band is equal to the (Max-Min)/2 ... so, "Average+/-Band" gives a pretty good idea of what's going on, without worrying about how many trades in the population. I'm providing this for: TradeQuality, TradeMDD, and #BarsInTrade

3. I'm providing a "sorta-Calmar" with note re "focus on max sustained drawdown", as well as the "HolmarkRR" with note re "focus on RTM, from Mark Holstius" ... thanks again, Mark!

4. The final set of three pairs of RwdRsk values are much easier to understand now ... Average of Trade NetPL/MDD, and Average of Trade NetPL/WTR each have their "Ngd" PctActiv variants ... then the *composite-risk* Average of TradeQuality values ... with Ngd variant. The TradeQuality "Band" from #1 pairs logically with the overall TQ values.

==========

Re the TradeLog engine ... I'm *finishing it up today* (hurrah). It is universal, very easy to use and understand, and doesn't require a bunch of file-maintenance by the user.
Top of the page Bottom of the page
Jump to page : 1
Now viewing page 1 [50 msgs/pg]
( E-mail a Link | Printer Version | Search Room )

Owner of site: Jim Dean -- Forum content is confidential, and may not be distributed without written permission.