For the Geeks

This page is for readers who will not trust the pane until they understand the shape of what is happening behind it. The goal is a mental model that lets you audit the tool's behavior on your own chart, not a recipe that lets you rebuild it from scratch. The distinction matters. A page that hands over formulas teaches a reader to replicate. This page is trying to teach a reader to interpret and to audit.

What this page will tell you

• The stages the tool runs through to turn a price series into a bounded pane.

• What each stage is shaped to do and what it costs.

• How the knobs on the configuration surface map to specific stages.

• Verification moves you can run on a 1-minute chart to confirm each stage is doing what this page says it is doing.

What this page will not tell you

• The exact transformation that maps an ATR-scaled distance into a 0..100 reading.

• The exact handling when ATR is zero or when the slow MA has not initialized.

• The exact rule for how missing slot values are absorbed inside the weighted blend.

• The exact internal ordering of clamp and smoothing steps beyond the high-level narrative below.

• Any detail sufficient for a reader to rebuild the script from this page.

Those protections are deliberate. The tool's value is partly in the careful shape of those details, and exposing them would give away the instrument without helping a reader use it better. If any question in the list above is one you genuinely need answered to make a decision, the mitigation is usually a verification on your own chart, not a look inside the code.

A useful way to read the rest of this page: treat each stage as a claim you can test. If a claim here is wrong about what a knob does on your chart, that is either a configuration you have not thought about (another knob is interacting) or a bug worth reporting. Either way, the right response is to narrow the test, not to assume the description is hiding something.

The stages, end to end

The pane is produced by a small pipeline. Each stage is a concept, not a line of code. What happens between the stages is protected; what each stage is doing is the shape you need.

Stage 1 — Raw distance

For each enabled slot, the tool measures the signed distance between the slot's chosen Source and the slot's baseline MA. The MA family and length are inputs on the slot. The distance is in the instrument's own unit (dollars, points, ticks, whatever the chart trades in) and carries a sign — positive when the source is above the baseline, negative when below.

What to take from this stage: the raw distance is an instrument-specific, unit-specific measurement. It is not comparable across symbols or across volatility regimes on the same symbol. If all the tool did was plot the raw distance, you would need a different set of reference lines for every ticker.

Stage 2 — Normalize by ATR

The tool divides the raw distance by an ATR value measured over the global ATR Length. That converts the distance into "units of recent volatility." A one-dollar distance on a calm instrument is different from a one-dollar distance on a violent one; after dividing by each instrument's own ATR, the two can be compared.

What to take from this stage: this is where the cross-instrument comparability is bought. The cost is that ATR itself has a personality. Short ATR windows hunt; long windows smooth out regime changes. The global dial is shared across all three slots — you cannot run a short ATR on the fast slot and a long ATR on the context slot. That is a deliberate simplification in the Base trim.

Stage 3 — Scale by sensitivity

The ATR-normalized distance is multiplied by ATR Sensitivity. Higher sensitivity pushes the scaled value toward extremes; lower sensitivity pulls it toward zero. This is a single scalar applied globally before the bounded transformation in stage 4.

What to take from this stage: sensitivity is a knob over where the bounded transformation starts biting. It does not change the underlying distance or the ATR scale; it changes where the clamp begins to matter.

Stage 4 — Bounded transformation into 0..100

The tool pushes the sensitivity-scaled value through a bounded transformation that maps any input to a value inside 0..100, with equilibrium at 50. The shape of this transformation is the part of the script that is protected. What you can count on as a reader:

• The mapping is smooth and monotonic within the resolvable range: larger distances map to larger readings.

• Above a certain magnitude, the mapping saturates — larger distances map to readings that are all near the bound. This is why pinning at 100 does not distinguish "very stretched" from "extremely stretched."

• The mapping is symmetric around 50 — the same distance above the baseline and below the baseline map to symmetric readings around equilibrium.

• The output is bounded defensively inside the 0..100 pane, even in corner cases where upstream inputs have behaved unusually.

What to take from this stage: the pane is bounded by design. The bound is the feature — it buys you comparability. It is also why readers have to stop treating 70 as "70% stretched" and start treating it as "in the part of the range that usually warrants attention."

Stage 5 — Slot slow line

The tool applies a smoothing MA (the slot's Slow Type and Slow Length) to the slot's fast value. The result is the slot's slow line. The fast and slow lines are both in 0..100 on each slot. The slow line is what the fast flips against to decide the slot's color.

What to take from this stage: the slow is smoothing an already-bounded series. The family choice matters less here than on the baseline; the length choice is what governs how often the slot changes color. Default 3 is intentionally short — a short-horizon regime read, not a long-horizon one.

During warm-up, the slot can pass through missing-value or fallback behavior before the Fast and Slow pair has enough history to be useful. Treat those early bars as warm-up, not as the settled color logic.

Stage 6 — Weighted blend across enabled slots

For the blend, the tool combines the enabled slots' fast lines into a single blended fast, and the slots' slow lines into a single blended slow, proportionally to the slot weights. Slots with weight zero are excluded from the average; slots with a missing fast value are handled conservatively. The result is clamped back into 0..100 defensively.

What to take from this stage: the blend is a weighted summary, not a vote. Equal weights produce an average; unequal weights tilt the summary toward the heavier slot. The exact weight normalization rule and the exact handling of missing slot values are not published; the verification move below lets you confirm the general behavior without seeing the rule.

Stage 7 — Optional master smoothing

If Enable Master Smoothing is on, the tool applies a final smoothing MA (Master MA Type, Master Length) to the blended fast and blended slow and re-clamps into 0..100. If master smoothing is off, the blend goes straight to the plot after the stage 6 clamp.

What to take from this stage: the master pass is a post-blend filter. It adds stability and adds lag. Both are real effects. The ordering matters: master smoothing runs after the weighted combination, not before. You cannot use master smoothing to "fix" a noisy slot before it enters the blend; the only way to reduce a slot's contribution to noise is through the slot's own settings.

How the knobs map to the stages

Quick reference so the configuration surface stops feeling abstract.

Verification moves you can run

Each of these confirms a specific claim this page makes, on a 1m chart of a liquid instrument, without needing to see the internals.

• Stage 1 and 2, together. Change the source on slot 01 from close to hl2. The slot's fast line should shift in a way that tracks the difference between the close and the midpoint of each bar, scaled by the instrument's ATR.

• Stage 3. Double ATR Sensitivity from 1.0 to 2.0 on the same chart segment. The pane should reach the 0 and 100 bounds visibly sooner. Halve it to 0.5 and the pane should compress toward the midline. The underlying price action has not changed; only the scale has.

• Stage 4. Push sensitivity to a large number (5 or higher). The pane should pin against the bounds for long stretches. That pinning is the clamp at work; it is confirmation that the tool respects the 0..100 bounds by construction, not by coincidence.

• Stage 5. Raise Slow Length on one slot from 3 to 20. That slot's color should flip much less often than the other two. The fast line's numeric range is unchanged; the slow line has just gained more memory.

• Stage 6. Set slot 02's Blended Weight to zero. The blend line should now track slots 01 and 03 only, and you can watch the blend's path change while slot 02's own line continues to draw. Then set slot 01 and 03 to zero as well; the blend should stop drawing entirely because there is no total weight to average.

• Stage 7. Enable master smoothing with Length = 10. Compare the smoothed blend to the unsmoothed blend on the same chart segment. Every bar where the smoothed version is a different color than the unsmoothed version is a bar where smoothing has delayed the regime read. That delay is the cost; you can see it directly.

• Cross-cutting. Toggle On Bar Close? between true and false. Slot lines should go from stable-between-HTF-closes (ON) to live-higher-timeframe behavior (OFF). Do not expect the two modes to match bar-for-bar in history; they answer different timing questions. See MTF & Repainting for the full verification.

The main tradeoff, named

A simpler design would have been a single-MA distance oscillator without the ATR normalization and the bounded transformation. That tool would be easier to explain. It would also mislead across instruments, volatility regimes, and timeframes. The Base trim's design chose a pipeline that:

• Makes reads easier to compare across instruments and sessions, by normalizing distance against each instrument's own ATR, while still leaving regime differences for you to judge.

• Stays bounded inside 0..100, by clamping the sensitivity-scaled value through a smooth transformation.

• Exposes the reader's curation to the reader — timeframes, baselines, weights — rather than hiding them in a preset.

The cost is configuration discipline. The reader has to own their choices, because the tool will not invent a "correct" configuration. This page is one piece of paying that cost without having to read the script.

Why the protected details are protected

A reader who has followed the pipeline to here has earned the question: why keep the exact transformation, the missing-value handling, and the weight-normalization rule out of this page? Three honest reasons:

• Reconstruction is not the same as interpretation. Nothing in the formulas would change how you read the pane on a live chart. The audit moves above let you verify the shape of every stage without the shape being published. Publishing the formulas would let someone rebuild the tool while leaving genuine users no better equipped than they already are.

• The protected pieces are where the careful work sits. Specifically, the sigmoid shape and its clamp behavior at the extremes, the handling when the ATR reference is unavailable or zero, and the way missing slot values are absorbed inside the weighted combination. Each of those is a small, careful decision that took real effort to get right. It is reasonable to keep the decisions, and not the reasoning about the decisions, inside the script.

• Interpretive clarity is the thing the pack is actually for. A reader who can describe what ATR Sensitivity does to the clamp and why pinning is not intensity will read the pane better than a reader who has the sigmoid equation memorized. The lines of investment this page argues for are the ones that improve interpretation. The lines that would only improve replication are the ones it holds back.

If the line between "protected" and "publishable" ever feels arbitrary, the test that keeps it honest is this: can a reader act better on a chart tomorrow with this information? If yes, it belongs on the page. If the answer is only "they could rebuild the tool," it does not.

Where to go next

• You have the shape and want to go back to decisions: Settings.

• You want the interpretive layer on top of this model: Visuals & Logic.

• You want the honest limits that come with this pipeline: Limitations & Trust Boundaries.