|Tools, Techniques, and Training for Nirvana Platforms|
|OmniTrader, VisualTrader, OmniFunds & OmniVest|
|Matthew 22:19-21 (ESV) … "Show me the coin for the tax.” And they brought him a denarius. And Jesus said to them, “Whose likeness and inscription is this?” They said, “Caesar’s.” Then he said to them, “Therefore render to Caesar the things that are Caesar’s, and to God the things that are God’s.”|
| Cumulative Volume TRAINING|
USA: GA, Lawrenceville
|This thread will collect examples from Questions I receive, with explanations as to why things happen as they do, when to the user, they may seem odd initially. Sometimes, the answers might suggest a possible need for tweaking the code to act differently, even though it's not currently "wrong". If so, I'll comment. Please post Q's on the Discussion thread, or via email ... I want to keep this thread "clean". |
NOTE: a lot changed - look at the final post first to see the newer graphics
USA: GA, Lawrenceville
|FIRST EXAMPLE: |
Salim sent me this case ... he wondered why the early-mid Sep price rise did not cause the bearish (red horiz line) TZ to end. He also noticed that the two histos immediately below that set of bars seemed "blank".
I responded that the line was "deep/dim" red, implying very few supportive bearish votes ... and that there were inadequate opposite-direction votes to kill the TZ. The "blank" histo's actually are correct ... all six curves are on the zero-line, which of course means that the histo-bar votes related to them are zero. The only potential non-zero vote might come from the middle Vmult line, which has a bit of green.
However, even that Vmult green "doesn't count" unless the CumV histo bar (see Detail screen) is above the FastMA enough (ie vs the Fuzzy gap) to confirm it as a "vote" (this is a nuance that's sort of cool, but not emphasized in the Help explanations).
HOWEVER ... if we reduce the Fuzzy input from its extreme value of 4, back to the default of 2 (the middle chart), then it allows one +1 bullish vote to appear as the price uptrend starts, from the Slope, and the other +1 vote that Ends the TZ (since the End votes input is 2) comes from the green Vmult where the histo CumV was sufficiently above its FastMA.
Re potential need for modification: this is working as it is supposed to. However, it's not clear from the Vmult middle line as to whether the CumV histo (on the Detail chart) is above, ~equal, or below the FastMA. That would have assisted with interpreting this situation. SO ... I am considering adding one more horizontal line, sort of like the middle one, along the very bottom. Its colors would indicate how the CumV histo value relates to the FastMA, considering the fuzzy setting.
Thanks for asking about this, Salim!
Later ... the Fix is In!
I decided to implement the mod described above, to visually clarify the two components that make up the +1/0/-1 BarVote (which, with Slope and Stack, allows for the +7 > -7 votes discussed in Help).
In doing so, I noticed that I'd failed to include "fuzzy" logic in the rule requiring the CumV histo to be adequately distant from the FastMA (above/below = bullish/bearish). So, I fixed that ... it impacts a few of the TZ results, but not many ... makes them smarter (ie sometimes it makes a current +1/-1 vote go away).
It looks good, and helps interpretation, to put the two lines together in the middle. The top one represents the Vmult magnitude (sort of like a slope, but not really). The bot one represents the Histo-FastMA spacing (just like stacking of MA's). So, the parallelism with the slope-histos above and stack-histos below makes it easy to remember which middle line is which.
Interpretation is simple ... if both lines are green or both are red on the same bar (any degree of brightness), then a +1 or -1 vote is created ... else the BarVote is zero. I'll need to update Help for this change. Note: keep on testing ... this will not have an significant influence on the conclusions you are drawing.
The final chart in the snapshot below shows this new logic and output in action ... I hope y'all like it!
(TRAINING- how fuzzy kills votes.png)
TRAINING- how fuzzy kills votes.png (196KB - 0 downloads)
USA: GA, Lawrenceville
|If you haven't seen the prior post re the (minor) BarVote calc change, and the (nice) improvement to the Plot=1 panel, be sure to check it out! This post relates to how that affects the Return values. |
Prior to this, there were 33 Return options. However, all we were "tracking" for the BarVote was the fuzzy-increments of the CumVmult ... which got zeroed out if the CumV histo was even a skosh below or above the FastMA in the opposite directiton of the CumVmult.
Now, however, the Histo vs FastMA "stack" is being evaluated like all the other stacks ... vs a fuzzy-increment. And its magnitude is being plotted as the new (bottom) horizontal line of the middle pair. So ... in line with all the other things, we need to have a parallel Return-value to report the same info that the chart does with its pretty colors.
To do this, I'm making the final digit of the Return= -1 output (the BarVote) be a magnitude that combines the CumVmult magnitude with the HstoVsFast magnitude, in a reasonable way. It's identifier (in Help) is now "B" instead of "S", and it has an output range of 1-9 (where 5= no vote, 1-4= bearish, and 6-9= bullish ... like other digits).
Unfortunately, that Return= -1 output value has maxed out the number of digits it can report (max=7) ... and to be consistent with the stack and slope votes, we really should track and report the *duration* of a given BarVote state ... so, to do that, AND to report the magnitudes and durations of the two components CumVmult and HstoVsFast, I need to ADD MORE return options ... seven more ... (and I remove one in the process since -31=PrcMult is now more sophisticated, in fuzzy-increments) ... bringing the new total to a nice round serendipitous 40 Return options.
The list will be:
-31 = Bb.MmSs ... where B= barV magnitude, M= multiplier, S= stacking ... all in fuzzy increments 1-9; b, m, s are durations of those states, much like t,p,k,a,b,c,x,y,s are for the other existing breakout components.
-32 to -36 are individual-digits b, M, m, S, s (you can get individual B already from option = -8)
-37 = Mm
-38 = Ss
... you can get Bb simply by using INT(result for -31)
-39 = Cumulative V actual value (used to be -32)
-40 = Capped V actual value (used to be -33)
I've just about gotten all the code in to do that. Obviously I also need to update Help for both the new Plotting and the new Return stuff ... and for the nuance to how the Voting is done.
OK ... that brings you up to date.
USA: GA, Lawrenceville
|The version A3 charts below represent what I expect is the final format for the four plot choices. |
Plot#1 (first chart) is relatively new ... it was in A2, but in A3 the middle "BarVote" line is split into top and bottom strips, showing the two components (price ratio multiple on top, and histo vs fastMA factor on bottom). Some small mod's were made to those components of the BarVote calcs, to bring them up to snuff re fuzzy calcs ... but it has a nearly negligible impact on the resulting TradeZones.
Plot#3 (second chart) - SLOPE VOTES - looks similar to A2's version, but the histo coloring is simpler (three shades not six), and denotes something different. The colors of each MA represent that MA's slope ... up, down, or fuzzy-flat. The colors of the CumV histograms represent the *net* of the Slope votes towards the TradeZone state ... green/blue/red = bullish TZ, neutral/inactive TZ, bearish TZ.
Plot#4 (third chart) - STACK VOTES - is also similar to A2's version, but again, the histo coloring is simplified and assigned differently.The colors of each MA represent one of the stacking votes ... fastMA > FvS stacking, medMA > FvM stacking, and slowMA > MvS stacking ... above, fuzzy-equal, or below. The colors of the CumV histograms represent the *net* of the Stack votes towards the TradeZone state ... green/blue/red = bullish TZ (+1 to +3), neutral/inactive TZ, bearish TZ (-1 to -3).
Plot#2 (fourth chart) - BAR VOTES - has been totally rebuilt for version A3. It uses five "layers" of histograms to convey six different items of information. It looks a little busy at first, but the colors make a lot of sense once you spend a couple of minutes studying it. Here's the story ...
1. the widest, dimmest background histo's show the Cumulative Volume magnitude, and their color represents the net BarVote value of green=+1=bullish TZ, blue=0= fuzzy-neutral, red=-1=bearish TZ.
2. on top of that, and *usually* not visible since it's covered up, is a skinny, med-blue histo spike (July 26 and Nov 16 are obvious cases) ... this is the "raw volume" value, before any capping logic is applied. When the RawV is under the cap-limit, these lines are hidden by the next "layer" ...
3. on top of those two layers, med-width, light-blue histos (always positive, above zero line) indicate the Capped Volume value ... which usually covers up the RawV unless it has a large spike. These histo's are skinnier than the dark CumV histos, so even on a bar where CapV is taller than CumV, you can see the color and height of CumV by looking at the outside edges.
4. on top of those three layers, a med-width red or green histo (same width as CapV histos) reports (like a thermometer) what portion of the CapV shares was the positive (lime) or negative (pink) change to that bar's CumV value. These are always plotted as positive values, so it's easy to see how much of the light-blue CapV histo was utilized. The percentage varies widely depending on the Cum Calc Method selected. For OBV, since the multiplier is always +1 or -1, the lime/pink histo's always completely obscure the light-blue CapV histo behind them, since they are always 100% of the shares.
5. all of the Cum Calc Methods are optionally affected (pretty significantly) by a Sliding Window of time ... that is, for any bar, the Sliding Window input dictates the number of prior bars that impact the net CumV value. This means that not only is a new percent-of-CapV added/subtracted to CumV at the RHS of the window ... but also a prior pct-CapV (from the LHS of the window) is backed out of the RHS CumV value. Thus the change in CumV at the RHS is the net of one new bar being included, and one old bar being removed. The #4 layer reports the sign and magnitude of the new-bar inclusion ... and this #5 layer, using the same math and colors, reports the sign and magnitude of the old-bar deletion, as a histogram plotted *below* the zero line.
Collectively, plots #2, #3, #4 provide visual detail/elucidation of the various components that are all summarized on plot #1. Note also that every useful piece of information that you can derive from these plots visually, is echoed by one or more of the 40 Return-value options.
Although the list of param's shows the Cum Method to be #8, I've since switched the param's back to the same order as the A2 version ... so A3 will show these charts, using VPT calcs, as the default #4 option.
Also ... the input-options 1-2 OBV calcs normally create huge CumV values (vs CapV), such that plot#2 would have tiny little light blue CapV bumps on the zero line ... so, I divide all OBV CumV's by ten (also for Return).
Similarly ... the input-options 3-4 VPT calcs normally create tiny CumV values (vs CapV0, such that plot#2 would have background dark-red/blu/grn histos that are so short you couldn't see them behind the tall CapV values ... so, I divide all the VPT CumV's by ten (also for Return) - fourth chart below shows how well that works.
The other CumV methods 5-8 don't have any x10 or /10 adjustments.
(QCOM plot-1, defaults (VOLeval A3); TradeZone focus.png)
(QCOM plot-3, defaults (VOLeval A3); SlopeVote focus.png)
(QCOM plot-4, defaults (VOLeval A3); StackVote focus.png)
(QCOM plot-2, defaults (VOLeval A3); BarVote focus.png)
QCOM plot-1, defaults (VOLeval A3); TradeZone focus.png (22KB - 0 downloads)
QCOM plot-3, defaults (VOLeval A3); SlopeVote focus.png (21KB - 0 downloads)
QCOM plot-4, defaults (VOLeval A3); StackVote focus.png (21KB - 0 downloads)
QCOM plot-2, defaults (VOLeval A3); BarVote focus.png (16KB - 0 downloads)
USA: GA, Lawrenceville
|Here is what I hope is the final set of revisions to the plot output formats. There have been a considerable number of changes, many thanks to Salim for his ideas and feedback. At this time, I'm uncertain as to which of these will be available in the "Core" version (leaning to just #1) ... all three will be avail in "Xprt". All the info on the available charts is echoed via the 40 Return-option outputs, either via several-items-in-one form, or in individual form ... if the Core version only has the first plot, then it will offer only 10-12 Return options. |
I'm trying to come up with a "sexier" name than "Volume Evaluation" ... needs a 3-letter abbreviation for the DLL name (like MTV). I'm thinking of "CVW" = "Cumulative Volume Wave". I'd refer to the "TradeZone" as the CVwave ... ie alluding to a "flow" of shares. Thus: CVWxprt & CVWcore. What do you think?
Plot #1 Top Band ... and its parallel return values (pane manually squished on purpose)
Band shows CVwave in nine colors that correspond to the (revised) initial "T" digit of Return option -1. Bright, medium and dim green/red show high,med,low #votes during a bullish/bearish wave ... wave must always start with BgnZon #votes of course. (T=9,8,7 bull and 1,2,3 bear). There are three "non-wave" colors ... darkgreen, navy, purple ... darkgreen and purple are "alerts" that a new Wave might soon begin (T=6,5,4). The "mapping" of votes to colors varies a bit, depending on what BgnZon input is.
BULL Colors: . . . Bgn7: 7, 6-4, 3-0; Bgn6: 7-6, 5-3, 2-0; Bgn5: 7-5, 4-3, 2-0
WAVE ReturnT: . . . . . . 9 . . 8 . . 7 . . . . . . 9 . . 8 . . 7 . . . . . . 9 . . 8 . . 7
... Bear Wave similar: 1,2,3 in place of 9,8,7 ... pure neutral is 5
... 6 is a "bull alert" (dark green) and 4 is a "bear alert" (purple)
Plot #1 Triple-Strips ... and parallel return values
Each strip shows the net *vote* for a category ... all vary from 2-8. For the upper and lower strips: 2-4 = 3,2,1 net Bearish Votes (red); 8-6 = 3,2,1 net Bullish Votes (green), and 5= zero net votes (navy) in that category. The upper strip shows SlopeVote, middle shows BarVote, and bottom shows StackVote. The middle BarVote strip has only one +/-1 vote (grn/red) ... so the fuzzy-increment-magnitude behind the vote is used to create alternate tints of the bull/bear colors ... anything green for that strip means +1 vote, and anything red means -1 vote.
Plot #2 is essentially unchanged from the major revision that is described in the prior post ... I tweaked the spacing a bit, and the mapping of colors for the upper CVwave band is as described above. But otherwise, the same. Note that this chart shows the net Slope/Stack Votes with the upper/lower histograms, and the +1/0/-1 Bar Vote is determined from the colors of the two middle strips ... if both are greenish (some tint), that's a Bull = +1 vote; if both reddish, that's a -1 Bear vote; in all other cases, it's a neutral "0" vote.
Plot #3 has been totally revamped ... it replaces all three of the prior 2,3,4 formats described earlier. Its purpose is primarily for training/explanations/understanding of "how" the inner calc's of the routine work ... it shows the "rudiments" of those calc's ... I do *not* expect people to use this for normal trading-decision purposes. The explanation in the prior post for plot #2 (with various layers of histograms) is an accurate starting point for this new plot #3. I've added the three Slope-Vote-colored MA's from the old plot#3 (see that prior explanation) to this chart. The new stacked "dots" provide the same information as the colors of the MA's in the old plot #4, indicating the variants in degree of Stacking, as described previously for the old plot #4. Yes, this can be sort of squished, depending on how the various components' values map to the y-axis ... but simply reducing the number of bars displayed, and either making the pane taller or maximizing the plot window, will always provide sufficient resolution.
(QCOM plot-1, defaults (VOLeval A3); TradeZone focus.png)
(QCOM plot-2, defaults (VOLeval A3); Components focus.png)
(QCOM plot-3, defaults (VOLeval A3); Rudiments focus.png)
QCOM plot-1, defaults (VOLeval A3); TradeZone focus.png (18KB - 0 downloads)
QCOM plot-2, defaults (VOLeval A3); Components focus.png (23KB - 0 downloads)
QCOM plot-3, defaults (VOLeval A3); Rudiments focus.png (26KB - 0 downloads)
|Owner of site: Jim Dean -- Forum content is confidential, and may not be distributed without written permission.|