Troubleshooting

Written By Axiom Admin

Last updated About 1 month ago

Troubleshooting

This page helps you diagnose problems with the indicator. It is organized by symptom — what you see on the chart — followed by the likely cause and what to do about it. Most problems are setup issues with straightforward fixes. Some are misunderstandings about what the tool is designed to do. And a few are genuine limitations that no configuration change can fix. The end of this page helps you tell the difference, because all three can look the same on the surface but need different responses.

Symptom-cause-fix table

Symptom	Likely cause	Fix
Runtime error on load	A slot's timeframe is lower than the chart timeframe. For example, Slot 01 is set to 5m but the chart is on a 15m timeframe.	Increase the slot's timeframe to be >= the chart timeframe, or switch to a lower chart timeframe.
Second runtime error after fixing the first	A slot's Fast Length is >= its Slow Length. The indicator requires Fast < Slow.	Set Fast Length to a value strictly less than Slow Length.
A slot line is completely missing	The slot is disabled (Enable toggle is off).	Open settings and check the Enable toggle for that slot.
A slot line disappears but the blend still reacts to it	The slot's Hide Plot toggle is checked. The slot runs and contributes to the blend, but its individual line is not drawn.	If you want the line visible, uncheck Hide Plot. If you want the slot's data in the blend without the visual, this is working as designed.
All three slot lines look identical	All three slots are configured with the same timeframe and same MACD lengths. With identical inputs, they produce identical outputs.	Check whether you changed the default timeframes (5m / 15m / 60m) to the same value. Restore different timeframes to see differentiated behavior.
The blended line tracks only one slot	The other slots are either disabled, have weight set to 0, or are returning `na`.	Check that all three slots are enabled and have non-zero weights. Also check that no slot is returning `na` (see below).
Values stuck near +100 or -100	ATR Sensitivity is too high. The oscillator saturates early, pushing readings toward the bounds.	Lower ATR Sensitivity toward 1.0 (the default). See Settings for details.
Values stuck near zero, oscillator seems unresponsive	ATR Sensitivity is too low, or ATR Length is too long relative to the data range. The oscillator barely reaches meaningful levels.	Increase ATR Sensitivity toward 1.0, or decrease ATR Length.
Oscillator is very jagged and noisy	ATR Length is too short, making the normalization denominator unstable. Or MACD lengths are too short for the chart timeframe, producing noisy raw MACD values.	Increase ATR Length (default is 14). Increase MACD Fast and Slow Lengths if they are well below the standard 12/26. Or use a slightly higher chart timeframe.
Slot line shows `na` or is missing at the start of the chart	Not enough historical data to calculate the MACD + ATR at the requested timeframe. This is normal warmup behavior.	Wait for enough bars to accumulate. The number of bars needed depends on the slot's timeframe and MACD/ATR lengths. Higher timeframes with longer lengths need more chart bars before they produce values.
Cross-ticker slot shows `na` or behaves oddly	The ticker symbol entered in Optional Ticker is invalid, does not exist, or does not have data at the requested timeframe.	Verify the symbol exists on TradingView and has data available at the slot's timeframe. Try typing the symbol in TradingView's symbol search to confirm its exact ID.
Blend jumps erratically	One or more slots are transitioning in and out of `na` state. When a slot enters `na`, the blend recalculates from the remaining slots, which can cause a sudden shift.	This usually happens during warmup or when a cross-ticker slot has intermittent data. Wait for all slots to stabilize. If the problem persists, check that all slot symbols have consistent data.
Alert did not fire when expected	Alerts are bar-close gated — they fire only after the chart bar closes. If the condition was true intrabar but reversed before bar close, no alert fires.	Check whether the condition was still true at bar close. If using On Bar Close = OFF, the slot values may have shifted between the intrabar condition and bar close.
Alert fired but the reading changed afterward	On Bar Close is OFF. The alert fired on a confirmed chart bar, but the slot values included unconfirmed HTF data that changed when the HTF bar closed.	If you need stable alert conditions, keep On Bar Close = ON (the default).
Regime flip alerts firing rapidly	K and D are very close together, causing frequent crossovers. This happens when the MACD and Signal lines converge — each small price movement can cause K to cross back and forth over D.	This is not a bug — it is a natural consequence of K and D being close. Consider increasing the Signal Length to smooth the D line, which reduces the frequency of near-equal crossings. Or ignore rapid flip alerts during periods of low momentum when K and D are converging.
Master Smoothing made the oscillator sluggish	Master Smoothing adds a final MA pass to the blended output, adding lag. The longer the Master Length, the more lag.	Reduce Master Length. Or turn Master Smoothing off if the lag cost exceeds the smoothing benefit. See Settings.
ALMA parameters seem to have no effect	ALMA Offset, Sigma, and Floor Offset only activate when a slot's MACD MA Type or Signal MA Type is set to ALMA, or when enabled Master Smoothing uses ALMA. If you are using EMA, SMA, or another non-ALMA type everywhere, the ALMA parameters do nothing.	Check which MA types your slots are using and whether Master Smoothing is enabled with ALMA. The ALMA parameters in PU Settings apply only to ALMA-based calculations.
A slot flatlines after choosing SWMA as the MACD MA Type	In the imported lite MA library, SWMA is fixed-length. Using it for both the fast and slow MACD legs removes the fast/slow distinction, so the MACD spread effectively collapses.	Pick a different MACD MA Type if you want a working fast/slow pair. If you need SWMA-style smoothing, remember it is fixed-length and does not honor the length inputs.

Deeper diagnostic reasoning

Some problems do not have a single line fix. Here is how to think through the less obvious cases.

"The oscillator looks right in history but wrong in real time"

This is almost always a repaint-related observation. If On Bar Close is OFF, the historical chart shows confirmed values (they look clean and stable), but the live edge shows provisional values that shift as the HTF bar builds.

What to check: Is On Bar Close set to OFF? If yes, switch it to ON and compare. If the historical view looks the same but the live view becomes more stable (values step-update on HTF bar close instead of shifting every chart bar), the issue is the expected repaint behavior of the OFF setting.

If On Bar Close is ON and you still see discrepancies, check whether you are comparing a different time period. Market conditions change — the oscillator's behavior depends on the data it is processing, so different periods can look different without anything being broken.

"The blend does not match what I expect from the individual slots"

This usually means the weight distribution is not what you think it is.

What to check:

Are all slots enabled? Disabled slots are excluded from the blend entirely.
Are all slot weights non-zero? A weight of 0 excludes the slot from the blend.
Is one slot returning na? Slots with na values are skipped in the blend calculation. If Slot 03 is na because it does not have enough data yet, the blend is calculated from Slots 01 and 02 only.
Is one weight much larger than the others? A weight distribution of 80 / 10 / 10 makes the blend essentially a copy of the dominant slot.

Try setting all weights to equal (33.3 each) and confirming the blend sits roughly between the three slot values. If it does, the blend math is working correctly and the issue is your weight configuration.

"The oscillator readings are different on the same instrument at different times of day"

This is expected. The oscillator is based on MACD and ATR, both of which respond to price movement. During active sessions (market open, news events), price movement is larger and more directional. During quiet sessions (pre-market, overnight, lunch hours), price movement may be smaller and noisier.

The oscillator is correctly reporting what the MACD and ATR calculations produce in those different conditions. Smaller price movement during quiet periods means the MACD spread is smaller, and the ATR may be adjusting to the lower volatility. This can make the oscillator look less responsive — not because it is broken, but because there is less momentum to report.

"I changed settings and now the oscillator looks completely different"

Some settings have a dramatic effect on the oscillator's character. The three most impactful are:

ATR Sensitivity — changing this from 1.0 to 2.0 can make the oscillator look like a different tool. See Settings.
Slot timeframes — changing from 5m/15m/60m to 1m/5m/15m changes the responsiveness of every slot.
On Bar Close — switching this changes whether you see confirmed or provisional data on the live edge.

If the oscillator looks dramatically different after a settings change, it is probably working correctly with the new settings. Compare the behavior against the expectations for those settings rather than assuming something broke.

Setup error vs. misunderstanding vs. genuine limitation

Not every problem is a setup error. Some problems come from misunderstanding what the tool is designed to do. And some come from genuine limitations that no amount of settings adjustment can fix.

Setup errors

These are configuration problems that have a concrete fix:

Slot timeframe below chart timeframe → increase the slot TF
Fast Length >= Slow Length → fix the values
Slot disabled or hidden when you expected it visible → toggle the setting
ATR Sensitivity far from default → return to 1.0 and re-evaluate

Misunderstanding

These are expectations that the tool was never designed to meet:

Expecting the oscillator to predict reversals from OB/OS levels
Expecting the blended reading to represent unanimous agreement (it is a weighted average)
Expecting historical stability when On Bar Close is OFF
Expecting the oscillator to generate buy/sell signals

These are not bugs. They are design boundaries. See Limitations and Trust Boundaries for the full picture.

Genuine limitations

These are things the tool cannot do regardless of configuration:

Resolve beyond ±100 when values are in the extreme zone (tail compression from the bounding function)
React faster than the ATR and MACD lookback periods allow (the lag is inherent)
Produce meaningful readings when the instrument has no price movement (MACD and ATR both need price activity)
Guarantee that cross-ticker normalization is equally stable for all instruments (volatile instruments produce noisier normalization)

If your problem falls into this category, the answer is not a settings change — it is understanding the boundary and working within it. That is not a dismissal. Knowing that the oscillator loses resolution at the extremes is useful information. Knowing that the lag is inherent lets you stop searching for the setting that removes it. The boundaries are not bugs to work around. They are properties of the math that the oscillator is built on, and working with them is part of using the tool seriously.