Visuals and Logic

This pane does a lot at once, and the reading discipline that keeps it honest is a specific ordered walk through what it shows. The order matters because the pane contains both fast reads (the slot and blend colors) and slower derived reads (the four structure features), and because the features are derived from the same blended line. If you start with a structure feature and work backward, you will usually be reading the pane out of sequence, and the thing you notice first will set the frame for everything else you read — including agreements that were built in by the construction rather than observed in the data.

The centerpiece of this page is the four-stage reading order. Install it here, use it in Workflows, and reinforce it every time you troubleshoot. The order is not a ritual; it is a sequence designed so each stage's read is calibrated by the stages before it.

The four-stage reading order

Read the pane in this order, every time, until the discipline is reflexive.

Stage one — slot color

A slot's line color encodes the slot's own RSI-vs-signal relationship. Full-tone color means that slot's smoothed RSI is above its own signal line. Faded (semi-transparent) means it is at or below. This is an RSI-vs-signal read, and the comparison is local to the slot.

It is worth holding this clearly, because the reflex from classical RSI is to compare the RSI line to 50. That is not what the color is reporting. The code includes a defensive 50-line fallback if a slot's signal value is truly unavailable, but the slot helper normally substitutes the current smoothed RSI until the MA pass can return a value. In that warm-up equality case, the line falls to its faded/down tone and no bullish or bearish state alert fires. The primary rule is RSI-vs-signal, per slot.

Slot colors are fixed: slot 01 teal, 02 aqua, 03 blue, 04 orange, 05 yellow. This is a design choice, not a limitation to apologize for — fixed colors make a five-slot stack legible across charts without the reader having to memorize a new palette each time. Line widths are per-slot if you want to emphasize or de-emphasize a slot visually.

Stage two — slot value

Where on the vertical axis the slot is sitting. The axis is native RSI, clamped into 0..100. Fifty is the midline. Seventy and thirty are reference bands that you can move. Zero and one hundred are structural rails the math cannot cross.

The reference bands are orienting tools, not triggers. No alert condition reads them; no color change keys off them; no structure feature inherits them. If you move seventy to seventy-five, the 70 line redraws at 75 and nothing else changes. This is an STR-specific thing to internalize early — the bounded pane looks like a classical RSI, and classical RSI habits want to treat seventy as a trigger. They are not that here.

Stage three — blend color

The blended RSI line's lime-or-red is a summary of the RSI-vs-signal relationship at the blend level — not a count of slot colors that agree. The blend runs through the same color rule as the slots: lime when blend RSI is above blend signal, red otherwise. A defensive 50-line fallback exists only if the blend signal is truly unavailable; ordinary reads are blend RSI versus blend signal.

This means you can sometimes see several slot colors pointing one way while the blend color points the other. That is not a contradiction; it is the weighted-mean math. The blend aggregates the slots' smoothed RSI values and their signal values separately, compares the two aggregates, and colors itself accordingly. A majority of slot colors leaning bullish does not force the blend color to be bullish if the weighted-mean blend RSI has just crossed under the weighted-mean blend signal.

If you need the blend to behave more like a vote count, that is a different tool. The blend here reports the conversation across the slots. It does not resolve disagreement.

Stage four — structure features, in sequence

Read the four features in this order: Donchian, Keltner, BBWP, divergence. Each sits on a separate axis of inspection; each says something slightly indirect about the blend line you just read. The ordering is deliberate — it runs from most directly geometric (Donchian's range edges) toward most derived (divergence's cross-pivot comparison), so the features that are easiest to anchor wrong are encountered last, when the slot and blend reads have already set the frame.

• Donchian. Silver steplines at the blend's own highest and lowest over the channel length. A Donchian press is the blend sitting at its own recent range edge by its own history. Not price's range edge. The blend's.

• Keltner. Orange envelope at the blend's own smoothed basis plus-and-minus the blend's range times the multiplier. A Keltner touch is the blend stretching against its own basis. Not the blend being overbought. Not the blend about to turn. Stretched against its own basis, by its own recent range.

• BBWP. Aqua-and-blue columns at the bottom of the pane. The columns rank the blend's own Bollinger-band width as a percentile of its own history. A tall column is the blend's width high against itself; a short column is the opposite. The qualifier "from the blend, not from price" rides every BBWP reading on this pane.

• Divergence. Small triangles in the pane's extremes when a confirmed chart-price pivot pairs with a blended-RSI value in the opposing-direction order at the same offset. Bullish triangle down low, bearish triangle up high. Geometry. Two pivots. Not a call on the next bar.

Each feature has its own visibility toggle on the settings page. Running all four simultaneously is allowed. It is not encouraged for a first read — the pane gets crowded, and the features' co-movement (treated at length below) is harder to see when they are all active at once. A good first-session move is to leave Keltner off (its default) until you have read the other three for long enough to know what each one is reporting on its own.

What each feature is actually saying, in plain language

Donchian — range-edge pressure, in blend-space

The Donchian feature wraps the blended RSI line with two steplines — one at the blend's highest over the channel length, one at the blend's lowest — and optionally a smoothed midpoint between them. When the blend prints a fresh extreme, the stepline steps to it. When the blend retreats, the stepline holds.

What this does not say: it does not say anything about price. The Donchian here is a Donchian of the blend. A Donchian press on this pane is not a price breakout.

Keltner — stretch, in blend-space

The Keltner feature places a basis (a smoothed blend line) at the center and wraps it with two bands, spaced by the blend's own bar-to-bar range times a multiplier. The bands describe how far the blended RSI has stretched from its basis, measured by its own recent range.

A Keltner upper touch means "the blend is stretched against its own basis." It does not mean overbought. It does not mean exhausted. It does not assert anything about price. Reading it as an overbought call is the most common Keltner misread on this pane and is named on Limitations and Trust Boundaries.

BBWP — regime of the blend's own width

The BBWP columns rank the blended RSI's current Bollinger-band width as a percentile of its own history across the lookback window. Classical Bollinger-width construction (matched-length basis and standard-deviation window), then a percentile-rank helper applied to the blend's widths.

Tall columns mean the blend's width is ranking high against its own history — regime of widening. Short columns mean the opposite — regime of compression. The column color flips across a user-configurable threshold (default 50.0).

The whole read is scoped to the blend. A BBWP-low column on a session where price is clearly moving is not a contradiction and is not a bug — it means the blend's width is low against its own history, which can happen during price-moves if the blend itself is not widening. The qualifier "from the blend, not from price" rides every mention on this page and every other page that mentions BBWP.

Divergence — confirmed-geometry between price pivots and blend

The divergence engine waits for a chart-price pivot to confirm — Pivot Len bars of price action on each side of the pivot — then compares the confirmed pivot to the prior confirmed pivot in the same direction. The bullish rule: chart price made a lower low at the confirmed pivot pair, while the blended RSI at the same offsets made a higher low. The bearish rule is the mirror.

When this comparison holds, a triangle prints. Bullish triangles at y=1 (near the bottom); bearish triangles at y=98 (near the top). Both evaluate only inside confirmed bars.

A triangle is geometry. It is a description of two price pivots and two blend values. It is not an entry. It is not a directional forecast. It is not a call on the next bar. Holding this qualifier tight is not a suggestion; it is the discipline this pane is built on.

The Plot On Pivot toggle changes where the triangle draws. Off (the default): at the right-shoulder confirmation bar. On: back-shifted by Pivot Len bars to the original pivot. The visual changes; the alert does not. The alert still fires on the confirmation bar regardless of the visual choice, and you did not have the marker at the back-shifted bar in real time. Pair both truths every time.

The co-movement of the four features — one line, four framings

This is the section that earns its placement. The four structure features read the same blended RSI line from four different inspection axes. They are not four independent witnesses. A BBWP compression, a Donchian press, a Keltner touch, and a divergence triangle at the same moment will look like four things agreeing. They are one thing reported four ways.

Hold that carefully. The construction does not add evidence by adding features. Adding Keltner to a pane that already has BBWP does not give you a second opinion; it gives you a second framing of the first opinion. The two framings are useful because they ask different questions — Keltner asks about stretch, BBWP asks about width regime — but their answers come from the same input. An agreement between them tells you that one line is showing one state from two angles. It does not tell you anything has been corroborated.

Three consequences you should feel, not just note.

First, apparent confluence is not evidence. When Keltner, Donchian, BBWP, and divergence all seem to point the same way, they are reporting the same line from four angles. It feels stronger than one feature alone, especially under time pressure, because the visual weight of four corroborating markers is hard to unsee. It is not stronger. If the line itself is wrong — misconfigured slots, duplication trap, mixed-posture blend, misread cross-ticker — the confluence inherits all of it, and the visual authority of the confluence is exactly the problem: the pane will look loudest about a reading that is least trustworthy.

Second, the language around this pane needs to refuse the word "confirms" between structure features. A divergence triangle does not confirm a Donchian press, and a BBWP reading does not confirm a Keltner touch. They co-exist. They share a source. Agreement between them is structural, not observational. The pack will not write "Keltner confirms BBWP" or "divergence confirms Donchian," and it asks you not to read them that way either. The verbs that honestly describe feature agreement are "co-exists with," "co-occurs with," "is consistent with." They are less satisfying. They are correct.

Third, if you want to feel the co-movement in your hands instead of just reading about it, run the drill in Workflows. Toggle one feature off and back on. Then change a slot's weight or enable state. The Keltner, BBWP, and Donchian reads move with the blend because they are built from it. Divergence recalculates from the same changed blend too, though you may not see a marker move at that moment unless the pivot geometry is present. That is the construction-level dependency on display, and once you have felt it you will stop reading feature agreement as breadth of evidence.

Two reads of the same pane

Picture a pane mid-session. The blend is green and near 60. Slot 01 is full-tone at 58. Slots 02 and 03 are full-tone near 65 and 62. Slot 04 is hidden with weight 0. Slot 05 is disabled. A bullish divergence triangle printed three bars ago. The Donchian lower stepline is a long way below the blend. The BBWP column is tall and aqua. Keltner is off.

A first-pass read under time pressure often sounds like this: blend is green and rising, divergence just fired, BBWP is tall, Donchian has room to the upside, four things agree, bullish.

None of those statements is factually wrong. What is wrong is the weight assigned to the agreement. The blend being above its signal is one observation about the blend. The divergence triangle is a geometric finding about a pair of price pivots and the blended-RSI values at the same offsets — it is not a forecast and it does not address the bar after the triangle. The tall BBWP column is a statement about the blend's own Bollinger-width regime, scoped to the blend, not to price. The Donchian stepline's distance is a record of where the blend has been, not a guarantee of where it is going. Three of those four features — Donchian, BBWP, divergence — are describing the same blended line from three framings. Two slots at 62 and 65 running the same source and length on different timeframes are also largely one reading sampled twice.

A mature read does not invert the first-pass read. It corrects for the weight. Same pane, same features, same colors: the blend is in an RSI-above-signal regime at a midrange value; the three-slot stack is largely coherent because the slots are configured similarly; the divergence triangle is a prompt to look at the two pivots rather than to act on the triangle; the BBWP and Donchian readings describe the blend's own history, not anything about the next bar or about price; the four-feature impression of agreement is a reflection of the fact that three of the four are the same line reported differently. The reader ends up with a smaller conviction than the first-pass read produced, and a more specific set of questions — is the slot configuration actually differentiating along meaningful axes, or am I reading a measurement I took three times; what did the two pivots look like that produced the divergence; would I accept the opposite read as easily if the blend color flipped on the next bar.

The difference between the two reads is not sophistication. It is whether the reader is letting the pane do their thinking for them. The pane will happily do it — visually, convincingly, on every session that looks tidy. The discipline that resists that is a walked sequence, every time, even when the pane looks easy.

A note on the tinted blend fill

The area between the blend's RSI line and its signal line is filled with a color matching the blend's own color at the moment — lime when the blend color is lime, red when it is red. The fill's hue is a state cue. The fill's extent (how much vertical space it takes up) is just the geometric distance between the two lines. A wide fill is not more bullish than a narrow one; it just means the two lines are further apart. Read the hue; do not read the area.

Where to go next

When you are ready to think about timing, repainting, and the per-slot bar-close control, MTF and Repainting is the next stop. When you are ready to wire alerts against what you have learned to read here, Alerts names everything the script fires on. When you want to see the four-stage reading order applied to four named scenarios, Workflows is where that teaching lives in scenario form.