Troubleshooting

Symptom-first triage for the most common confusions on STR. Find the symptom closest to what you are seeing, check the likely cause, run the quick check, apply the fix. If the symptom is not here, it probably belongs on Limitations and Trust Boundaries — the difference is that limitations are things the tool is designed to do that look wrong; troubleshooting is things that actually are wrong or that hide a setup mistake.

The two pages exist separately because they demand different reading postures. Troubleshooting is "what is broken, and how do I fix it fast." Limitations is "what does the tool refuse to do, even when correctly configured." A reader who tries to troubleshoot a limitation will lose time; a reader who tries to accept a setup error as a limitation will lose more. When you are stuck, ask yourself whether the behavior is unexpected (probably troubleshooting) or expected and uncomfortable (probably limitations). Wrong first guess costs ten minutes. Not asking costs days.

Every entry below carries: symptom, likely cause, quick check, fix or next step.

Category A — setup errors

These are configuration mistakes the tool tells you about, usually through Pine's runtime error channel.

Symptom: A runtime error message reading MACD 0N timeframe cannot be lower than the chart timeframe.

Likely cause. A slot's Timeframe input is below the chart timeframe (e.g., slot set to 5m on a 15m chart).

Quick check. Open the Inputs dialog, check the Timeframe of the slot named in the error. Compare to the chart TF.

Fix. Raise the slot's TF to at or above the chart TF. Or set it to empty string (which means "run on the chart TF" and collapses the slot to the chart bar).

Symptom: A runtime error message reading MACD 0N Fast Length must be less than Slow Length.

Likely cause. The slot named in the error has a Fast Length greater than or equal to its Slow Length.

Quick check. Open the Inputs dialog. Look at the slot's Fast Length and Slow Length.

Fix. Make Fast strictly less than Slow. The classical pair is 12 / 26 but any pair where Fast < Slow is valid.

Symptom: The indicator added successfully but the pane is empty — no lines, no columns, no triangles.

Likely cause. Every enabled slot is returning na. Common reasons: a very new symbol where the slots haven't warmed up yet; a source that doesn't exist on the slot's timeframe; every enabled slot has weight 0 (no blend draws).

Quick check. Scroll backward a few hundred bars. Do plots appear on older bars? If yes, the current bars are still warming up for your lookback settings. If no, check slot Enable flags and weights.

Fix. Wait for history to fill in, or ensure at least one enabled slot has non-zero weight, or change the source to a series that exists on the selected symbol and timeframe.

Symptom: BBWP columns are blank at the left edge of the chart, for what looks like months of history.

Likely cause. BBWP Lookback defaults to 252 bars. BBWP returns na until the percentile-rank window has 252 prior valid width readings, and the width calculation itself has to warm up first.

Quick check. Count the bars from the left edge of the chart to the first BBWP column. If it is roughly the lookback plus the Bollinger-width warm-up, the column is not broken.

Fix. No fix needed. Scroll forward in history. If you need BBWP earlier, reduce Lookback at the cost of a less-robust percentile rank (and note that 252 gives a rough one-year ranking window on a daily chart after warm-up; shorter lookbacks produce noisier percentiles).

Category B — behavior that looks wrong but is the tool being honest

These items read like bugs until you understand what the tool is doing. They are not actually bugs. Each one is behavior the design calls for, named here so you can recognize it without thinking it is broken.

Symptom: A slot line is missing from the pane.

Likely cause. One of: (a) the slot's Enable is OFF; (b) the slot's Hide Plot is ON; (c) the slot's K is still na (warm-up); (d) the slot's source series is na on the slot's timeframe; (e) your own zoom has that bar outside the draw window.

Quick check. Open Inputs, check Enable and Hide Plot. If both are "right," scroll right to see if the slot appears on recent bars.

Fix. If Hide Plot is ON and you want to see the line, flip to OFF. If Enable is OFF and you want the slot contributing, turn it ON. If warming up, wait. Remember: Hide Plot ON still leaves the slot voting in the blend.

Symptom: The blend moved, but no visible slot's line moved.

Likely cause. A hidden slot (Hide Plot ON) is moving and pulling the blend, or a visible slot's line moved below your visual threshold and you missed it.

Quick check. Temporarily set Hide Plot OFF on all enabled slots. Watch the next bar or two. The moving slot will reveal itself.

Fix. No fix needed. If you want to monitor all contributing slots visually, leave Hide Plot OFF on the ones you want visible. If you want a clean pane, keep Hide Plot ON but remember hidden slots still pull.

Symptom: A per-slot alert fired, but the slot's line is hidden on the pane.

Likely cause. Hide Plot controls visibility only. Alerts are driven by the slot's computed state (K > D or K < D), regardless of whether the line is drawn.

Quick check. Open the Alerts panel. Confirm the alert is for a hidden slot.

Fix. No fix needed. If you want alerts to fire only when the slot is visible, add a downstream filter on your side. The indicator does not gate alerts on visibility.

Symptom: A per-slot alert fired, but that slot's weight is zero.

Likely cause. Weight controls blend contribution only. A weight-zero slot still computes its own state and fires its own per-slot alerts.

Quick check. Open Inputs, check the slot's weight.

Fix. If you want the slot fully silent, disable it (Enable OFF). If you want it as an observer (plots and alerts, no blend vote), weight zero is correct.

Symptom: The divergence alert fires late — sometimes ten or twenty bars after the visual pivot.

Likely cause. Pivot Len (default 20) sets the right-shoulder confirmation delay. A divergence is not confirmable until the right-shoulder pivot has closed Pivot Len bars after the pivot.

Quick check. Count the bars between the right-shoulder pivot (the higher low in the second leg of a bullish divergence, or the lower high in a bearish one) and the current bar. It should be close to Pivot Len.

Fix. No fix — this is the cost of strict pivots. Lower Pivot Len for faster confirmation at the cost of more false pivots. Higher for fewer and later. There is no best value.

Symptom: Divergence triangles visually appear to shift backward after the bar they printed on.

Likely cause. Plot On Pivot is ON. The setting back-shifts the triangle by Pivot Len bars to land on the original pivot bar.

Quick check. Open Inputs, Blend Div group, Plot On Pivot?.

Fix. If ON is what you want (you want to see where the geometry formed), keep it on and remember the alert fired on the right-shoulder confirmation bar, not on the visually-anchored bar. If you want the triangle to stay on the alert bar, set Plot On Pivot OFF.

Symptom: All MACD Slots Bullish (or Bearish) alert fires on most confirmed bars of a session.

Likely cause. If slot 05 is disabled, your slots may not be differentiated. Most often: same source, same MA family, same length pair, differing only in timeframe. The alignment alert fires whenever a single convergence-divergence story holds across repeated cadences — which, in a trending session, can be most bars.

Quick check. Open Inputs and look at each enabled slot's source, MA family, and length pair. Count how many are identical.

Fix. Differentiate deliberately. Vary sources (close / hl2 / hlc3), vary MA family on at least one slot (ALMA or Jurik, say, for the 60m slot), or vary length pair. If your process is specifically built on matched-slot consensus reading, accept the repeat-fire and use the alert as presence rather than event.

Symptom: Every visible slot is aligned, but the All MACD Slots Bullish or All MACD Slots Bearish alert does not fire when slot 05 is enabled.

Likely cause. Current source overcounts active slots inside the slot 05 alignment branch. That makes the all-slots comparison unreliable when slot 05 is enabled.

Quick check. Disable slot 05 and test the same alert behavior on slots 01..04, or read alignment visually instead of relying on the built-in all-slots pair.

Fix. No settings fix inside this source version. Treat the all-slots pair as unreliable with slot 05 enabled until the source path is corrected.

Symptom: A slot's line sits against 100 (or 0) for many bars.

Likely cause. ATR Sensitivity combined with the current volatility regime has pushed the sigmoid into its flat region. The slot is saying "the convergence gap is large relative to recent ATR," and the sigmoid's saturation is eating additional magnitude.

Quick check. Temporarily set ATR Sensitivity to 0.5 and watch the slot's line retreat toward the midline.

Fix. Reduce Sensitivity or increase ATR Length if pinning is a frequent session behavior. Do not act on pinning as intensity — 100 is an asymptote, not a reversal zone.

Symptom: Master smoothing is ON and the per-slot alerts still seem noisy.

Likely cause. Master smoothing applies to the blended K, D, H triple only. Per-slot alerts fire off per-slot states, which are not affected by master smoothing.

Quick check. Open Inputs, confirm Enable Smoothing is ON. Then note which alerts you are seeing — per-slot or blend?

Fix. For noisier per-slot alerts, adjust the individual slot's length pair or MA family. Master smoothing is blend-scope only.

Symptom: BBWP reports a low column on a visibly volatile price chart (or a high column on a quiet one).

Likely cause. BBWP is the blended MACD line's own Bollinger band width percentile. It is not price volatility. A loud price session can still have a compressed blend if the slots disagree and the weighted mean flattens.

Quick check. Compare BBWP to a price-BBWP indicator if you have one loaded. They will differ.

Fix. No fix — BBWP on STR is blend-derived by design. Read it as a framing question about the blend's own width regime, not a volatility read on price.

Symptom: The pane's color theme looks wrong (e.g., slot lines all one color).

Likely cause. The slot color settings under the plot-definition section have been overridden. By default, slot 01 is teal, 02 aqua, 03 blue, 04 orange, 05 yellow, with full tone above the slot's own signal and faded below.

Quick check. In the Style tab of the indicator's settings, verify the per-slot color values match defaults.

Fix. Reset to defaults or set your preferred per-slot colors.

Category C — genuine limits, not fixes

Some things cannot be troubleshot; they are genuine boundaries of the tool.

Symptom: You want divergence per slot, not only on the blend.

Limit. STR ships divergence on the blend only. Per-slot divergence is deliberately not in this trim.

Next step. If ten-slot context breadth is the requirement, CTX is the surface to evaluate. If divergence around the blended MACD line is the requirement, that is STR's job. No single version should be described as both CTX's ten-slot context surface and STR's blended-line structure surface.

Symptom: You want alerts on Keltner touches, BBWP threshold crossings, or Donchian channel edges.

Limit. STR does not carry structure-feature alerts. The pack's position is that alerting on non-independent structure features invites the "one line agreeing with itself" misread. You can build downstream flip detectors on the blend's exported plots (which are visible in the data window) if you want such triggers on your own side.

Next step. If the missing alert is a dealbreaker, build it downstream from the exported plots or choose a version whose alert surface actually matches the job.

Symptom: You want more than five slots, or ten slots with STR-style structure features.

Limit. STR caps at five because the structure layer consumes processing budget. Its structure reads live on the blended MACD output, not on each individual slot.

Next step. Use CTX when ten-slot context breadth matters most. Use STR when blended-line structure matters most. If you need ten slots and STR's structure layer at the same time, that is outside this family split.

Symptom: You want three slots without the structure features — just the clean normalized pane.

Limit. STR carries the four structure features whether you use them or not (though you can toggle them all off). The pane will still be drawn as STR.

Next step. Base, which is three slots plus the blend and histogram and nothing else. If you are running Base territory, use Base; it will not feel like a demotion.

Symptom: You want to remove the ATR-sigmoid normalization and see raw MACD magnitudes.

Limit. The ATR-sigmoid transformation is the identity of this pane. Removing it would turn STR into a different tool.

Next step. If raw magnitudes are what you need, use TradingView's built-in MACD in addition to (or instead of) STR. Some readers keep a classical MACD on the chart alongside STR for both reads.

When the tool and your reading disagree

If the pane is showing you something that does not match your read of price, the pane is not wrong by default and your read is not wrong by default. Both can be true, because the pane is describing one thing (the convergence-divergence story across your chosen slots) and your price read is describing another. The useful question is not "which is right" but "what is the pane telling me that my price read is not catching, and vice versa?"

That is the whole point of a context tool. If you find yourself wanting the tool to confirm what you already believe, you have stopped using it as a context tool and started using it as a mirror. The pack's position is that mirrors are expensive — they cost money when they agree with a bad idea, and they cost learning when they keep you from noticing that your reading and the pane are telling genuinely different stories.

A practical move when the disagreement feels sharp: turn off three of the four structure features, read the bare pane (blend, histogram, slot lines if un-hidden) against price for ten minutes, then turn the features back on one at a time. Often the disagreement is the reader processing four feature layers at once and conflating them. Stripped down, the blend either disagrees with price (informative) or it does not, and the feature layers come back into a reading routine that knows what the base layer was saying.

If you are still stuck, the combination of Limitations and Trust Boundaries and For the Geeks often resolves the residual confusion — one teaches what the tool is designed to do, the other teaches why the transformation and the features behave the way they do. Most residual-confusion questions get answered by one of those two pages.