Troubleshooting
Use this page when the pane is behaving differently from what you expected.
Written By AxiomCharts
Last updated About 2 hours ago
Troubleshooting
Use this page when the pane is behaving differently from what you expected.
The fastest way to solve most problems here is to separate three questions:
- is the slot setup legal
- is the slot active, hidden, or weight-zero
- is the stack carrying a timing assumption I forgot about
Do not troubleshoot and interpret at the same time. Get the stack back to something legal and explainable first. Then decide what the chart is telling you.
Start with the symptom you actually have
When the problem is not obvious, use this rescue order:
- fix timeframe legality first
- confirm which slots are enabled, hidden, and weight-zero
- confirm timing posture
- remove alternate-symbol influence until the same-symbol stack makes sense again
Symptom: the script errors immediately when I load it
Likely cause
At least one enabled slot timeframe is below the chart timeframe.
The shipped baseline is `5 / 15 / 60`, so this usually shows up when the chart is above `5m`.
Fix
- open settings
- inspect every enabled slot
- raise the conflicting `TimeFrame:`, clear it so it inherits the chart timeframe, or disable the slot
- reload your reading only after the chart is legal again
Verify
You should be able to say which enabled slot now has the lowest timeframe on the chart.
Until you can say that, do not move on to smoothing, alerts, or blend interpretation.
Symptom: the blended pair is missing
Likely cause
The blend has no valid weighted contributors.
That usually means:
- every enabled slot has `Blended Weight:` set to `0`
- or the weighted contributors are unavailable because the stack is not configured the way you think it is
Fix
- inspect the active slots
- find at least one enabled slot with a valid value and positive weight
- confirm `Plot Blended K/D` is on if you also want the visuals visible
Verify
Set one active slot from `0` weight back to a positive value and confirm the blended pair returns.
Symptom: I hid a slot, but it still affects the stack
Likely cause
`Hide Stoch NN Plot` changes visibility only.
The slot can still:
- calculate
- trigger slot alerts
- affect alignment
- affect the blend if it still has positive weight
Fix
- If you want the slot gone, turn off `Enable Stoch NN`.
- If you want the slot active but quiet, keep it hidden and make sure you understand whether it also needs weight or alert participation.
Verify
Hide a slot and then disable it. Compare what changed in each step. If those feel different now, the problem is solved.
Symptom: the blend and alignment do not match
Likely cause
The blend is weighted. Alignment is counted.
A slot with weight `0` can still count toward alignment. A heavily weighted slot can dominate the blend without creating full-stack agreement.
Fix
- list the enabled slots
- list which of those have positive weight
- compare that list against the slots alignment is still counting
Verify
Set one enabled slot to weight `0` and watch whether alignment still includes it while the blend stops using it.
Symptom: the slot moved while I thought it was confirmed
Likely cause
`On Bar Close?` is off for that slot, or you are watching a different slot than the one you thought was settled.
Fix
- open the slot's power-user section
- check `On Bar Close?`
- if needed, compare that slot against a duplicated confirmed version
Verify
Run a side-by-side confirmed versus live-forming comparison on one higher-timeframe slot.
Symptom: the alert did not fire until the bar closed
Likely cause
All alerts in this script are evaluated on chart bar close.
Even if the slot is using live-forming requested-context values, the alert surface itself is still close-gated.
Fix
- confirm which alert family you used
- confirm the chart timeframe you are watching
- adjust your expectations around close-based evaluation instead of assuming intrabar behavior
Verify
Watch one state alert and one event-style alert on replay. Confirm both still wait for chart bar close.
Symptom: the stretch lines feel wrong
Likely cause
You may be reading `70` and `-70` as if they were classic stochastic `80 / 20`.
They are not.
They are stretch markers on this tool's own centered scale.
Fix
- reset your expectations around the tool's scale
- move `Overbought Level` or `Oversold Level` only after you know why you are moving them
- compare the centered pane against a standard stochastic instead of forcing direct equivalence
Verify
Change only the stretch lines and confirm that you changed a tool-defined attention zone, not the underlying slot construction.
Symptom: another symbol made the stack harder to trust
Likely cause
The alternate-ticker slot may be:
- weighted into the blend before it earned that role
- carrying a different timing posture
- adding a story you never clearly assigned it
Fix
- isolate the alternate-symbol slot
- keep it visible
- set its weight to `0`
- compare it against the same-symbol baseline before deciding whether it belongs in the blend
Verify
If you can explain the slot's job in one sentence after that reset, the stack is getting healthy again.
Symptom: master smoothing made the pane calmer but less useful
Likely cause
That is the real tradeoff.
Master smoothing happens after the blend exists, so it can make the summary look cleaner while delaying flips, threshold touches, and midpoint changes.
Fix
- toggle `Enable Master Smoothing` off
- compare the smoothed and unsmoothed blend side by side
- keep the calmer version only if the delay still fits the workflow
Verify
Watch at least one regime-flip or threshold event with smoothing on and off before making it permanent.
Quick reset path
If the stack has become too hard to explain, reset to this:
- three active same-symbol slots
- confirmed mode on all three
- equal positive weights
- no alternate ticker
- no master smoothing
- visible slot lines
That baseline is not the final answer. It is the shortest path back to something you can reason about again.
If the problem is really trust, not setup
If the settings are legal and the pane still feels more persuasive than clear, go here next:
Visual placeholder: Troubleshooting table screenshot with callouts for timeframe legality, hidden-versus-disabled behavior, weight-zero blend loss, and alert timing.