Troubleshooting

Written By Axiom Admin

Last updated 22 days ago

Troubleshooting

Organized by symptom category, because the useful skill when something looks wrong is triage. Before you fix anything, decide what kind of problem you have. Is the configuration rejecting itself, is the pane working correctly and reporting something easy to misinterpret, or is the tool working as designed and the behavior you wanted is simply not wired here? The three categories below correspond to those three questions, and the categorization matters because the remedy differs — a setup error wants a configuration change, a misread wants a reading-discipline change, and a genuine limit of the tool wants either a workaround elsewhere in your workflow or an acceptance that the tool does not do the thing you wanted.

Each row follows the same shape: symptom, likely cause, quick check, fix or next step. The quick check is the cheapest thing you can do to distinguish this row from neighboring rows; run it before you commit to a fix. If the quick check does not distinguish, read neighboring rows — several symptoms on this pane can look similar at first glance and diverge on close inspection.

Setup errors

These are the symptoms where something about the configuration is rejecting itself at load or at use.

Symptom	Likely cause	Quick check	Fix or next step
Runtime error: `"<slot label> timeframe cannot be lower than the chart timeframe."`	An enabled slot's timeframe input is below the chart timeframe. Often happens after switching the chart to a higher timeframe and leaving a slot where it was.	Open the indicator settings and compare each enabled slot's timeframe against the chart timeframe.	Raise the slot's timeframe to at or above the chart timeframe, or lower the chart timeframe to at or below the slot. There is no third option — the script refuses lower-than-chart slot timeframes.
A slot plots nothing at all, even with `Hide Plot` off.	The slot's `Enable` is off, or the slot's timeframe is empty and the chart has not accumulated enough bars to populate the slot's smoothing and signal passes.	Check `Enable` for the slot. If on, check whether the slot's timeframe is empty (in which case the slot runs on chart cadence) and whether the chart has enough history.	Enable the slot if it is off. If the slot is running on chart cadence (empty TF), wait for warm-up.
Power User parameters have no visible effect.	The MA family selected for the pass does not use the parameter you changed. For example, Jurik Phase does not affect ALMA.	Check the slot's `RSI Type` (or `Signal Type`, or structure-feature basis type). The Power User block covers multiple families; only the parameters matching the selected family are active.	Either change the family to one that uses the parameter, or tune the parameter that corresponds to the active family.
A cross-ticker slot is returning nothing.	The ticker string is invalid for TradingView, or the exchange is inaccessible on the current chart's data subscription.	Verify the ticker resolves on a standalone chart.	Correct the ticker string. If the exchange is inaccessible, the slot will not compute; either use a different benchmark or upgrade the data subscription.

Misreads

These are the symptoms where the pane is working correctly and reporting something that is easy to mis-interpret. The fix is usually a reading-discipline correction, not a settings change.

Symptom	Likely cause	Quick check	Fix or next step
BBWP columns are missing for a long stretch near the left edge of the chart.	The percentile-rank helper returns nothing until the full lookback window (default 252) is populated with prior blend widths. A fresh chart or a sparsely-populated instrument has not yet accumulated enough blend-derived widths.	Check the BBWP `Lookback` value. Consider how many bars have to accumulate before the window is full.	Wait for the chart to accumulate enough history. For debugging, temporarily lower `Lookback` to see the columns appear earlier — then restore the value you actually want.
A bullish divergence triangle printed and price did not turn upward.	Divergence is confirmed geometry between two chart pivots and two blended-RSI values. It is not a directional call or a forecast.	Re-read the divergence-as-question card in Workflows and the divergence section on Limitations and Trust Boundaries.	Nothing to fix in the tool. The expected posture is to treat the triangle as a prompt to inspect the underlying slot stack, not as an entry.
The blend is green and rising but a divergence triangle printed bearish.	The divergence engine compares chart price pivots to blended-RSI values at the same offsets. Directional disagreement between chart-price trend and blended-RSI pivot geometry is precisely what the engine reports.	Look at the two pivots that produced the triangle and compare with the blend's pivots at the same offsets.	Nothing to fix. The triangle is a question about the pivot pair, not a call to reverse the blend.
The `All RSI Slots Bullish` alert never fires when slot 05 is enabled.	Upstream current behavior — the alignment counter's slot-05 arm contains additional increments past the intended single increment, which inflates the active-slot count beyond what the per-slot bullish count can ever reach. As a consequence, the alignment alerts cannot fire when slot 05 is enabled.	Disable slot 05; confirm the `All RSI Slots Bullish` alert begins firing under appropriate conditions. Re-enable slot 05; confirm it stops firing.	Workaround until upstream remediation: treat the all-slot alignment alerts as four-slot-maximum signals. Do not rely on them when slot 05 is enabled. Per-slot, blend-state, and divergence alerts are unaffected. See Alerts for the context block and Limitations and Trust Boundaries for the framing.
The blend wobbles intra-bar more than expected.	Mixed posture in the blend: some slots at `On Bar Close? = true`, others at `false`. The `false` slots contribute live values that repaint until their HTF bars close; the blend inherits the mix.	Check each enabled slot's `On Bar Close?` in the Power User block.	If the mix is unintentional, set every enabled slot to the same posture (usually `true`) and compare. If the mix is intentional, read the structure features with the mixed-posture caveat in MTF and Repainting.
Structure features look jumpier than they did yesterday.	Something about the blend's input changed — slot enable state, weight, timeframe, source, or posture — and every structure feature over that blend inherits the change.	Compare today's slot configuration against yesterday's. Toggle one slot at a time to feel which change moved the structure features.	Reconcile the configuration back to what you expected, or accept the change deliberately. The structure features are not independent of the blend; they move when the blend moves.
A per-slot bullish or bearish alert keeps firing every confirmed bar during a sustained regime.	All alerts on this pane are state alerts, not transitions. A sustained condition fires the alert on every confirmed bar where the condition holds.	Check Alerts for the state-vs-transition framing.	For flip-only firing, do the edge detection on the receiving side (your routing system, your journal). The script does not carry transition alerts.
A BBWP column is tall while the chart looks quiet (or short while the chart looks busy).	BBWP ranks the blended RSI's own Bollinger-band width against its own history. The measurement is scoped to the blend, not to price.	Re-read the BBWP section on Visuals and Logic and the BBWP row on Limitations and Trust Boundaries.	Nothing to fix. The column is reporting the blend accurately. The qualifier "from the blend, not from price" applies.
A Keltner upper-band touch is happening and nothing is turning down.	The Keltner envelope on this pane wraps the blended RSI's smoothed basis. A touch means stretch against the blend's own basis; it does not mean price overbought or an incoming reversal.	Re-read the Keltner section on Visuals and Logic and on Limitations and Trust Boundaries.	Nothing to fix. Keltner reports stretch, not exhaustion.
I hid a slot to declutter and the blend is behaving in ways I cannot explain from the visible slots.	`Hide Plot` does not disable a slot. The slot still computes, contributes to the blend, and fires its own alerts as long as `Enable` is on.	Open the indicator settings and check `Enable` state for every slot.	If you want the slot truly off, disable it. If you hid it for legibility and want to keep it active, accept that you cannot read its contribution from the pane alone.
I set a slot's weight to zero and I am still getting its alerts.	Weight zero removes the slot from the blend only. The slot still computes and still fires its own alerts.	Check the slot's `Enable`.	Disable the slot if you do not want its alerts.
Plot On Pivot on back-shifts the triangle visually, but the alert fired at a later bar.	That is the intended behavior. Plot On Pivot changes where the triangle draws; the alert always fires on the confirmation bar (Pivot Len bars to the right of the original pivot).	Re-read the Plot On Pivot treatment on MTF and Repainting and Alerts.	Nothing to fix. The marker and the alert carry different timing claims — the marker can be backward-looking when ON; the alert fires on confirmation.
The divergence engine is firing triangles in every direction at once.	Pivot Len has been lowered to chase "more" divergences. Lower Pivot Len relaxes pivot strictness, producing more pivot pairs and therefore more triangles — without making any of them stricter.	Check Pivot Len. The default is 20.	Restore Pivot Len to 20 or higher. Pivot Len is the strictness dial, not the sensitivity dial.
Master smoothing is on, the blend looks cleaner, but alerts feel late.	Master smoothing adds lag to the blend by design. Blend-state alerts fire later, not earlier. Per-slot alerts are unaffected.	Check the blend smoothing `Enable`.	If the lag is a problem, turn master smoothing off. If the calmer plot is what you wanted, accept the timing cost.

Genuine limits of the tool

These are the behaviors where the pane is working exactly as intended and the gap between what you want and what the tool does is a design choice.

Symptom	Likely cause	Quick check	Fix or next step
No alert exists for the blend crossing 70 or 30.	Threshold-crossing alerts are not wired in the current script.	Check Alerts; the reference bands are visual bumpers only.	Build the threshold trigger on the receiving side if your workflow depends on it, or use a separate tool.
No alert fires on the bar a slot first crosses above its signal.	The script only carries state alerts, not transitions. A bar where a slot first becomes bullish fires the same alert as every subsequent bar where the slot stays bullish.	Same page as above.	Do edge detection on the receiving side, or combine two consecutive state reports into a flip signal in your own routing.
No alert fires before a divergence's right shoulder confirms.	Both divergence alerts gate on confirmed pivots. Live-pivot divergence detection is not wired.	Check Alerts and the divergence timing section on MTF and Repainting.	Accept the strict-pivot cost. A strict pivot confirmed late is usually preferable to a loose pivot detected early.
No alert fires on combinations like "Keltner upper touch while Donchian press while BBWP above threshold."	Cross-feature alerts are not wired. The four structure features co-exist as independent visibility toggles; nothing combines them at the alert layer.	Same page.	This composite is not available from the current script's alert surface. Build separate custom logic outside this alert list if your workflow needs it. The hidden `bullCount` / `bearCount` plots only carry slot-state count context; they do not expose Keltner, Donchian, or BBWP conditions.
No feature exists to restrict alerts to a specific session.	The script does not carry session-windowing on alerts.	Check Alerts.	Filter by session at the receiving side.

When to ask for help rather than keep tweaking

There is a specific pattern that this pack wants you to catch in yourself, because it is one of the most common ways a reader loses a session to a pane that was fine. Stop tweaking and take a step back if any of these are true:

You are on your fourth or fifth configuration change in a single session and you are no longer sure what problem you were solving. At that point you are almost certainly worsening the configuration rather than improving it, regardless of which direction you are moving the knobs.
You are enabling or disabling features to produce a read you have already decided you want to see. That is not using the pane to read the market; it is using the pane to validate a pre-existing conclusion, and the pane will oblige for the wrong reasons.
You are reducing Pivot Len or raising slot weights to get a signal you are missing. The tool does not owe you a signal, and tuning toward a feeling usually costs more than it saves. If the pane is quiet, the market may be quiet; if the pane disagrees with you, the disagreement may be the signal.
You are about to describe a structure feature using the word "confirms" in the pack or in your own notes. That word is a sign that the co-movement truth has slipped out of working memory, and the read is drifting.

In those moments, close the indicator settings, re-read the four-stage reading order in Visuals and Logic, and let the pane report for a few minutes before touching anything else. You will usually find the thing that confused you resolves itself into a reading question rather than a configuration question. If it does not, stop the session — a tired reader making configuration changes is not the same person who reads the pane carefully, and the difference between those two readers is where live-account damage happens.