Limitations and Trust Boundaries

Every tool has an estimate boundary — the line past which the tool stops knowing things and starts being misused. This page draws STR's.

A reader can configure this pane correctly and still misread it. That is the critical fact that shapes this page. Most of STR's over-trust damage is not at the settings layer; it is at the interpretation layer. Readers who follow the quick-start walk, verify their setup, and land on a defensible starter configuration can still lose real money by treating a triangle as an entry or by trusting an alignment alert that is really self-agreement.

This page leads with those two highest-damage misreads because if the reader absorbs nothing else from the pack, those two are the ones worth taking. The rest follow in descending order of damage potential. None of them are disclaimers. They are the estimate boundary, named plainly, so a reader can keep the pane in scope.

The two misreads that cost readers the most

Misread 1 — divergence triangles read as entries

A confirmed bullish divergence triangle is two confirmed price pivot lows in lower-low order paired with two blended-MACD values at the same historical offset in higher-low order. That is a geometric fact about a specific moment. The triangle does not predict direction, does not obligate price to turn, and does not survive into the next bar by default. The bearish version is the mirror.

Readers lose money on this misread because the triangle is the most visually magnetic element on the pane. It prints on exactly the kind of moment when a reader is already tempted to fade price. A reader who has been waiting for "confirmation" can treat the triangle as confirmation and act. The pane has been honest — the alert fired, the triangle printed on a confirmed pivot pair, the geometry is exactly what the module claimed. The reader imputed a directional claim the module never made.

The honest use of divergence. A confirmed triangle is the moment to ask whether the slot stack underneath the blend agrees with the triangle or argues with it. If slots are still confidently bearish under a bullish divergence, the triangle is a geometry note inside a continuation — interesting, not actionable. If slots are rolling over, the triangle is geometry that agrees with evidence already in the stack — a useful piece of context to factor into an already-forming view. Either way, the triangle is input to judgment, not a call by itself.

Every page on this pack carries the same qualifier. This page names it hardest because this is where the cost is largest.

Misread 2 — alignment read as breadth

The visual alignment read asks how many enabled slots have K above D, and how many have K below D, on a confirmed bar. The built-in All MACD Slots Bullish and All MACD Slots Bearish conditions are intended to report unanimous enabled-slot alignment, but the current source overcounts active slots when slot 05 is enabled. Until that source path is corrected, do not treat those two alert conditions as reliable for five-slot alerting. The visual misread still matters either way: "every slot agrees" reads like five independent voices reaching a consensus.

It is not, unless the slots are differentiated. If your five slots share source, share MA family (all EMA, say), and share length pair (all 12/26/9), and only differ in timeframe, visual alignment is five cadences of the same measurement. When all five agree, you have learned one thing, not five. If you use the built-in alignment pair after the slot-05 source issue is corrected, that alert will still need the same skepticism because the underlying observation is mostly self-correlated.

STR makes this trap more costly than Base, not less. With three slots a duplicated-observation alignment looks like three-way consensus. With five, the same alignment looks like five-way consensus and feels more convincing. The trap scales with slot count.

How to not fall in. Before you trust an alignment read, ask: do my slots look at different sources, run through at least one non-EMA family, include at least one cross-ticker, or use distinct length pairs? If yes, alignment is more meaningful. If no, it is self-agreement reported on a confirmed bar.

This misread and the divergence-as-entry misread together account for most of the honest mistakes readers make with STR. Put them at the top of your attention.

The rest, in descending order of damage

Boundary pinning read as intensity

A slot's K sitting against 100 for several bars is not telling you the convergence gap is twice as extreme as a reading at 80. It is telling you the ATR-sigmoid transformation, under the current Sensitivity and volatility regime, has pushed the sigmoid into its flat region, where different scaled values produce similar bounded outputs. Zero and 100 are asymptotes, not reversal zones.

The operational danger is two-sided. One: a reader fades the boundary pinning as "overbought" or "oversold" and takes a mean-reversion trade against a momentum regime that can run for days. Two: a reader extrapolates "more intense" when the line has already saturated, and sizes up on a move the pane cannot actually describe beyond "large enough to saturate the sigmoid." Both readings import magnitude information the bounded pane does not carry at the extremes.

The fix is usually to reduce ATR Sensitivity or increase ATR Length until the slot operates in the sigmoid's responsive region for most of the session. The fix is not to act on the pinning as if it were a signal.

BBWP is not price volatility

BBWP columns look like a familiar price-BBWP indicator. They are not. They are a percentile rank of the blended MACD line's own Bollinger band width. A low column means the blend has been in a compressed width regime by its own history. A high column means the blend has been in an expanded width regime by its own history.

A price chart can be ripping while BBWP reports low — because the blend, being a weighted mean of normalized slot reads, can compress when the slots disagree. A price chart can be quiet while BBWP reports high — because the blend's band width can expand on a steady move.

Every time you see BBWP, think "from the blend, not from price." The qualifier rides every mention on this pack for a reason.

The four structure features are not independent

Divergence, Keltner envelope, BBWP, and Donchian channel all read from the blended fast line. They are four framings of one underlying line. When a divergence triangle prints, the blend is often near a Keltner or Donchian extreme, and the BBWP has often just moved. That is not four independent features confirming each other. It is one line agreeing with itself by construction.

Reading two structure features as two agreeing witnesses is a misread that inflates confidence. The pack does not use confirms to link structure features and neither should the reader.

When the features do move together, that is consistent with the mechanics. When they disagree — e.g., a Donchian upper touch while BBWP is compressed — that is informative about the blend's shape (the blend is printing a fresh high but its own width is not expanding). The useful reads come from noticing how the features differ, not from stacking them as agreement.

Four-state histogram misread as regime

The four-state histogram is a speedometer, not a regime classifier. Bright green does not mean "established uptrend"; it means the blended histogram is above 50 and rising compared to the previous bar. One bar above 50 with one bar of up-movement is enough to color a column bright green. A single bright green column does not describe a regime. It describes acceleration at a specific bar.

Read the histogram as continuous evidence — how often is it bright vs faded, how long has it been on one side of 50, how symmetric are the transitions. That is regime reading. A single column's color is a speedometer reading, not a summary of the session.

Keltner touch read as overbought or oversold

The Keltner envelope wraps a smoothed basis of the blend with a bar-to-bar range estimate on the blend, scaled by a user multiplier. A touch means the blend is stretched against its own recent range. It does not mean exhaustion. In a trending regime, the envelope will be touched repeatedly — that is what a stretched regime looks like on a stretch envelope. The envelope is not a fade signal and it is not an OB/OS framing.

Master smoothing read as a reliability upgrade

Master smoothing applies to the blended K, D, H triple after blending. Turning it on makes the blend calmer. Blend-based alerts fire later, not earlier. Per-slot alerts are unaffected.

If you turned smoothing on because blend alerts looked choppy, be honest with yourself: the chop was evidence. You have hidden it, not solved it. Smoothing is a legitimate tool when the raw blend is reacting to noise you want averaged out — but it is a lag-for-stability tradeoff, not a "truer" read.

Per-slot On Bar Close? read as a magic switch

STR moves On Bar Close? from a global control (Base) to a per-slot control. Per-slot control is not free. It is more expressive — you can treat HTF slots and chart-TF slots differently, or treat two HTFs differently if you have a specific reason. It is also more error-prone: a reader who flips one slot to OFF without thinking through how it interacts with the others is producing a blend whose repaint posture is mixed across its members.

The switch is a tradeoff between lag (ON) and repaint (OFF) at each slot. It is not a solution to the repaint problem. Treat every OFF as a specific choice you can describe the reason for.

Chart-timeframe collapse

A slot with Timeframe empty runs on the chart TF. That is not wrong but it removes the MTF benefit for that slot. If you want the chart TF as one of your five voices, collapse is deliberate. If your slots are collapsing because you forgot to set TF strings, that is a misconfiguration.

Short-slow slot noise

A slot configured with a very short slow length (e.g., 5) on a noisy timeframe will produce a whippy normalized read. The slot's K will spike toward 0 or 100 often, the slot color will flip frequently, the per-slot alert will repeat-fire. If your slots are firing on almost every bar, check the slow length. A slow length below 10 on a 1m chart is usually contributing noise, not signal.

Cross-ticker slot surprises

Optional Ticker lets a slot read a different symbol's source and ATR. Useful for benchmark voices (e.g., a slot on SPY while trading a single-name stock). Also a source of blended-read surprises: if the benchmark moves while the traded name does not, the blend moves for reasons the chart's own bars cannot explain. That is a legitimate use case (the benchmark is supposed to influence the blend) but be aware that the pane is no longer describing a single symbol.

The tool does not repaint — a claim the pack refuses to make

A common request from readers evaluating tools of this shape is a "does it repaint?" answer. The honest answer for STR is: it depends per slot. A slot with On Bar Close? ON returns the previous slot-bar value and is steadier on confirmed history. A slot with On Bar Close? OFF can repaint until the slot bar closes. A blend whose slots are mixed across ON and OFF carries the repaint posture of its weaker members. Plot On Pivot back-shifts the divergence marker visually but does not change the alert's bar of record.

The pack will not tell a reader the tool does not repaint. That would be a partial truth presented as a full one.

What STR does not do

• It does not fire trade triggers. The alerts commit to state and geometry, not to action.

• It does not predict direction. The divergence triangle is geometry, not a forecast.

• It does not describe price volatility. BBWP is the blend's own width, not price's.

• It does not resolve disagreement between slots. The blend averages disagreement; the average is not a conviction.

• It does not replace price action. It is a context tool that adds to your reading, not a replacement for it.

• It does not cover every reader's needs. Three slots and a cleaner pane live at Base. Ten-slot cross-asset context breadth lives at CTX. STR is the structure version at the end of the family: five slots because the extra processing budget goes into blend-framing structure layers.

Reader's responsibility

The pack's job is to explain consequences. The reader's job is to make choices. Every configuration decision on STR has a cost named in both directions. Every structure feature is a question the reader asks, not an answer the pane gives. Every alert is state or geometry, not a trigger.

A reader who treats this pane as a conversation partner about what their MACD stack is doing will get real work out of it. They will notice when the blend is averaging disagreement instead of reporting conviction, they will notice when a divergence triangle agrees with slots already rolling over versus prints inside a continuation, and they will develop a sense of when the structure features are saying something new versus repeating what the blend already showed.

A reader who treats this pane as an oracle will be disappointed — not because the pane lied, but because the pane never promised to answer the question the reader was asking. The pane answers "what is the convergence-divergence story across my chosen slots, and how does the blend's own structure frame it?" It does not answer "should I take this trade," and pretending otherwise is the single largest source of losses readers report back on tools of this shape.

If this page's framing feels blunt, that is deliberate. The cost of sugar-coating the estimate boundary is a reader who mistakes honest instruction for a promise the tool never made. The tool is adaptable, configurable, and auditable on its own surface. It is not infallible. Keep those two facts in the same hand and the pack will age with your reading.

Troubleshooting specific symptoms: Troubleshooting. The construction of the four structure features at mental-model altitude: For the Geeks.