Workflows

Written By Axiom Admin

Last updated 22 days ago

Workflows

This page describes three named workflows for Axiom CVD Osc STR. Each one is a way of using the tool that the pack endorses — meaning the controls, the reads, and the timing all line up with how the indicator is built. Each workflow has a place where it breaks, and those places are named explicitly in an anti-pattern list at the bottom of each workflow rather than buried elsewhere. If a workflow looks appealing to you because it promises a read the tool is not designed to deliver, the anti-pattern list is the right section to check.

These are not the only ways to use STR. They are the three that match its design center. If you are already using STR in a way that does not appear below, the absence is not a rejection; it is that the pack wants to be concrete about what it has tested in mind, rather than generic.

Workflow 1 — Context stack with structure framing

What this is for

Building a working chart read where the default three-slot stack — chart cadence, one step up, one regime up when you start from a 5-minute chart — is the core pressure read, and the four structure features around the blend are framing on top of that read. This is the design-center workflow for STR. If you are not sure which workflow to try first, it is this one.

What this is not for

Building ten-slot context breadth. Use the CTX trim for that.
Generating entry signals from the structure features. None of them are entries.
Treating Keltner and Donchian as independent reads that can confirm each other. They share a source.

Setup

Leave slots 1, 2, and 3 at defaults (5m, 15m, 60m, equal weights).
Leave slots 4 and 5 disabled.
Keep Hide Plot ON for slots 1–3 until you specifically want to see them underneath.
Leave BBWP, Donchian, and divergence ON (defaults).
Turn Keltner ON once you have watched a few sessions with the default three-feature configuration. The fourth feature is additive to the read; adding it before you know what the other three are saying will crowd the pane.
Leave Plot On Pivot OFF.

How to read

The read runs in three passes every time you look at the pane. With practice it takes a few seconds. New, it can take a minute. Both are fine; the goal is the sequence, not the speed.

Pass 1 — what is the blend doing? Where is it in the pane — near 50, at an extreme, crossing its signal? What is the fill telling you about the recent relationship between the blend and its signal? If the blend is near 50, carry one question into Pass 2: agreement or disagreement underneath? If the blend is near an extreme, carry a different question into Pass 2: is one slot carrying the move, or is the stack genuinely agreeing? The answer changes how much weight the blend's position deserves.

Pass 2 — what is the slot stack doing? Unhide the slots if you have been running with Hide Plot ON. Look at the three lines. On the default 5-minute starting chart, identify the fastest one (slot 1, chart cadence), the middle one (slot 2, one step up), and the slowest one (slot 3, one regime up). On any other chart timeframe, first check that the slot timeframes still form the stack you think they do. Is the fast slot doing all the work? Is the slow slot late to agree, or has it not turned at all? Are they in broad agreement, with the spread between them narrow? A blend near 50 with agreeing slots is genuinely neutral. A blend near 50 with disagreeing slots is contested pressure — the slots are saying opposite things and the blend is summarizing the disagreement at its midpoint. Those are different pane states, and only the slot stack can tell them apart. A reader who skips Pass 2 will systematically misread the blend at exactly the moments where the read matters most.

Pass 3 — what is the framing saying? Look at the structure features in order. Is the BBWP column low or high, and has it just flipped across the threshold? Is the Donchian channel expanding or pinching? Is the blend pushing the upper or lower stepline? If Keltner is on, is the blend stretched on one envelope line, or sitting near its basis? These four reads describe the blend's behavior against its own recent history. They do not describe price. They are framing — they tell you which kind of blend behavior you are looking at, not what to do with it.

After Pass 3, you know where the pressure read is, whether the stack agrees or disagrees, and what the blend's own recent regime has been. That is the read. It feeds into whatever decision process you already trust for the instrument; it does not replace that process. If the read at the end of Pass 3 contradicts what your decision process was about to do, that is information — not a vetoer, but a question worth answering before you commit. If the read corroborates what your decision process was about to do, that is also information, but smaller than it feels — most of the time the pane and the chart are telling you variations of the same story, and the corroboration is the floor, not the proof.

What to watch for

The blend leading the slot conversation. Sometimes the blend will print a clear direction while one or two slots lag. That is normal — the blend is a weighted mean; a strong fast slot will pull the blend before a slower slot catches up. The question is whether you want to read the blend as leading or as out ahead of its stack.
Regime transitions in BBWP. When BBWP flips across the threshold in the same direction the blend is already moving, that is the blend's own width regime changing during a directional move. Worth noticing. It is not a trigger.
Donchian resetting during a quiet stretch. A channel that pinches on a quiet blend says the blend has been going nowhere by its own history. That is a statement about pressure going still, not about price.

Anti-patterns

Enabling slots 4 and 5 with default weights before you have a specific question they answer. The blend becomes the average of five contexts you did not curate.
Running all four structure features on a small chart with all three slot lines visible. Visual overload has a cost that is easy to underestimate. If the pane is too crowded to read in three seconds, cut something.
Reading Keltner-plus-Donchian agreement as confirmation. They share a source.
Turning Master Smoothing on early. It will calm the blend before you know what the blend is trying to show you.

Workflow 2 — Divergence as a question

What this is for

Using the divergence triangles as a structured prompt to re-read the rest of the pane, not as an entry. The divergence module is the most magnetic visual on STR. If your instinct is to act on the triangle itself, this workflow is the correction.

What this is not for

Trading the triangle as a turn call.
Scanning the chart for "strong" divergences by lowering Pivot Len to increase frequency.
Using the divergence alert as an automated trigger.

Setup

Use the same base setup as Workflow 1.
Confirm Show Div is ON and Plot On Pivot is OFF.
Set Pivot Len to match the swing scale you actually read. Defaults (20) are reasonable for intraday on a 5-minute chart. Higher lookbacks produce fewer, stricter triangles and longer confirmation delays. Lower lookbacks produce more, noisier triangles.

How to read

When a triangle appears:

Identify the confirmation bar and the pivot bar. With Plot On Pivot OFF, the triangle is on the confirmation bar. The pivot itself is Pivot Len bars earlier. Know where both sit before you do anything else. If you cannot point at the pivot bar on the chart within five seconds, you do not yet have your bearings on the geometry.
Look at the slot stack at the pivot bar. Not at the confirmation bar — at the bar the geometry is describing. Were the slots agreeing or disagreeing at that moment? Was the fastest slot already reversing, or was it stubbornly trending with price? A divergence whose pivot bar coincides with a slot stack already turning is a different read from a divergence whose pivot bar sat in the middle of a strongly trending stack.
Look at the blend's context at the pivot bar. Was the blend stretched on the envelope (if on) or mid-range? What was the BBWP read? Was the Donchian pinching or expanding? Hold the answer in mind without weighting it as confirmation — these features all read from the same blended line, and the divergence triangle is a fourth reading of that same line.
Compare the two confirmed pivots that define the divergence. The triangle describes the geometry between them. What was the chart doing during the span between those pivots? A divergence across a span where price consolidated is a different read from a divergence across a span where price trended hard, and a different read again from a divergence across a span that contained an exogenous shock. The span context changes how much weight the geometry deserves.
Ask whether the rest of the pane corroborates the geometric story. If the slot stack was already shifting at the second pivot, if the blend was already bending, if BBWP was expanding on the second pivot — the divergence is a description that fits the rest of the read. If none of that was happening, the divergence is geometric coincidence inside a different kind of market. Both are real states. The first is the state you can do something honest with; the second is the state that catches readers who acted on the triangle alone.

At step 5, you have not made a decision; you have finished reading the geometry in context. Whether to act on the read is a separate question that belongs to your process. The reading is not the trade. The reading is the work that lets you know which trade — if any — your process should consider.

A useful drill: do this entire reading pass on the next ten triangles you see, and write the result down — what the slot stack was doing at the pivot bar, what the blend's context was, what happened in the next twenty bars after the confirmation bar. Do not act on any of them. After ten such readings, you will have the start of your own honest sense of when the geometry is worth leaning on for your instrument at your timeframe. That sense does not arrive from reading this page. It arrives from doing the readings.

What to watch for

Confirmed divergences that print into continuations. These are common. A triangle is not a turn. If the rest of the pane was not corroborating at the pivot bar, the continuation was always the higher-probability read.
Clusters of triangles at low Pivot Len. Multiple triangles in a short stretch at a low lookback are not multiple confirmations. They are the same noisy region being described several times under a too-loose pivot definition.
Divergences across session resets. A bullish divergence whose second pivot sits shortly after a reset is describing geometry on a blend that had just restarted its accumulation. Treat it with extra care.

Anti-patterns

Lowering Pivot Len to "see more divergences." More triangles is not more information.
Trading the triangle and ignoring the slot stack.
Using Plot On Pivot ON in your working configuration. The back-shifted marker will train you to expect a real-time signal that does not exist.
Routing the divergence alerts into an automated executor.

Workflow 3 — Alert triage, not alert triggers

What this is for

Using STR's alert surface as a queue of observations to review, rather than as a stream of triggers to act on. This is the intended working pattern for all sixteen alert conditions.

What this is not for

Automating entries from any alert.
Wiring the blended bullish or bearish alerts directly into order execution.
Treating All CVD Slots Bullish as a confluence trigger.

Setup

Pick a small set of alerts that match what you want to watch. A useful starter set: All CVD Slots Bullish / Bearish, Blended CVD Bullish Divergence / Bearish Divergence, and one per-slot alert on whichever slot matches your execution cadence (usually slot 1).
Route the alerts to a place where they accumulate rather than act. A notification stream you can review at the top of each hour, a notebook entry, a dashboard column — anywhere that gives you a checkable log rather than a moment of pressure.
If you need webhook-style routing, send the alerts to a review queue, never to an order endpoint.

How to read

At review time, not at alert time:

For each state-descriptor alert, ask: what was the blend doing at that bar? What were the slots doing? Was the alert early, late, or roughly on time for the move you now see in hindsight?
For each divergence alert, run the Workflow 2 reading pass in review mode. The Plot On Pivot honesty implications are the same whether you read the alert live or after the fact.
Look for patterns across many alerts. Are the per-slot bullish alerts on slot 3 (the slow one) systematically arriving late to moves that slot 1 had been firing on for a while? That is a regime statement, not a parameter problem.

The value of this workflow is the distance it puts between the alert and the action. Pressure-state alerts in real time invite over-reaction; pressure-state alerts reviewed later invite learning.

What to watch for

Repeat firings during sustained states. A slot that has been bullish for thirty confirmed bars has fired thirty Is Bullish alerts. That is the alert working correctly, not a malfunction. If the repeat volume is overwhelming, either throttle in your alert platform or add edge detection on the receiving side; the script does not do that for you.
Alignment alerts shortly after a reset. All CVD Slots Bullish firing in the first bars after a daily session reset is often reset-noise. Mark those in the log so you do not over-weight them.
Divergence alerts outside your swing scale. A bullish divergence confirmed across a Pivot Len that is much longer than anything your process acts on is a geometry event at the wrong altitude. It is not wrong; it is simply not actionable inside your current frame.

Anti-patterns

Wiring any alert to an executor.
Treating the blended bullish alert as confluence with a slot bullish alert. They are different summaries of the same data; they are not independent confirmations.
Assuming the divergence alert is a trigger because it is one-shot. One-shot means "it does not repeat." It does not mean "act on it."
Adding more alerts than you can review on the timescale you actually review on.

Where to go next

For the alert inventory and the posture behind every firing rule, Alerts.
For reading the pane in the sequence Workflow 1 and Workflow 2 describe, Visuals and Logic.
For the confirmation timing that makes the divergence alert one-shot, MTF and Repainting.
For the honest limits that matter when you scale a workflow to a real process, Limitations and Trust Boundaries.