Alerts

The alerts on STR report state, not triggers. Every one of them gates on barstate.isconfirmed, which means every alert fires on a confirmed bar and never on tick. The state-holding alerts will fire on every confirmed bar where the state holds, not only on the bar where the state flipped. If you wire a state-holding alert straight to an executor, expect a flood during any sustained regime. If you want flip-only behavior, add the flip logic downstream of the alert.

Nothing here fires a trade. Some of the alerts are seductive — the all-aligned and the divergence alerts in particular — and the page names that out loud next to each one. Treat the alert surface the way you would treat a radio call from someone you trust to report what they see but not to decide what you should do about it. The alert names the state. The decision stays with you.

The sixteen conditions, grouped

Five groups. Ten per-slot conditions, two blend-state, two all-aligned, two divergence. No threshold crossings. No transition-edge events. No per-slot divergence. What is absent is deliberate and is named at the bottom.

Per-slot (ten total)

For each of slots 01 through 05, two conditions:

• MACD 0N Is Bullish — fires when slot N is enabled and slot N's K is above slot N's D on a confirmed bar.

• MACD 0N Is Bearish — fires when slot N is enabled and slot N's K is below slot N's D on a confirmed bar.

These fire regardless of whether the slot's line is visible on the pane. Hide Plot does not gate the alert. They also fire regardless of the slot's weight in the blend. A weight-zero slot still fires its per-slot alerts. The only thing that gates a per-slot alert is the slot's Enable flag.

What they commit to. State, not transition. On a sustained bullish slot state, the Is Bullish alert will fire on every confirmed bar where the state holds.

Blend state (two)

• Blended MACD Is Bullish — blend K above blend D on a confirmed bar.

• Blended MACD Is Bearish — blend K below blend D on a confirmed bar.

These are the blend-level mirror of the per-slot conditions. They fire on every confirmed bar where the weighted mean of the enabled, non-zero-weight, non-na slots is above or below its own signal. If every enabled slot is warming up or every active weight is zero, the blend is na and neither alert fires.

Master smoothing affects these directly. With smoothing ON the alerts fire later than they would without it. Per-slot alerts are unaffected by master smoothing.

All aligned (two)

• All MACD Slots Bullish — every enabled slot has K > D on a confirmed bar, and at least one slot is enabled.

• All MACD Slots Bearish — every enabled slot has K < D on a confirmed bar, and at least one slot is enabled.

Alignment is intended to count enabled slots only. Hide Plot and weight do not affect the intended count. A current source defect matters, though: when slot 05 is enabled, the active-slot total is overcounted, so the built-in all-aligned pair should not be treated as reliable in five-slot configurations until that source path is corrected. Visual alignment still works; the caveat is specific to these two built-in alert conditions.

The alignment trap. Unanimous alignment reads like breadth. "All five slots agree" sounds like five independent observations arriving at the same conclusion. The alert does not commit to that reading.

Whether the alert means what it sounds like depends on slot differentiation, which the alert does not police. If the slots are differentiated (different sources, at least one non-EMA family, at least one cross-ticker, distinct length pairs), "all five agree" is closer to what the alert sounds like it reports — five reads from measurably different vantage points agreeing. If the slots share source, MA family, and length pair and only differ in timeframe, "all five agree" is one measurement sampled at five cadences agreeing with itself, and the alert will fire on most confirmed bars of any trending session for almost no informational gain.

STR makes this trap more expensive than Base did, not less. A three-slot unanimous alignment looks like three-way consensus; a five-slot unanimous alignment looks more convincing by one-and-two-thirds slots' worth of confirmation bias. The extra slots feel like more evidence. They are not, unless you made them into more evidence by differentiating them.

Alignment alerts earn their weight through slot differentiation, and the pack cannot verify that for you. Configured lazily, they are the most seductive and least informative alerts in the set. Before trusting one, first keep the slot-05 source caveat in mind; then glance at the slot configurations in the inputs dialog and ask whether this alignment is convergence across voices or self-agreement across cadences.

Divergence (two)

• Blended MACD Bullish Divergence — a confirmed bullish divergence geometry has just evaluated on a confirmed bar.

• Blended MACD Bearish Divergence — a confirmed bearish divergence geometry has just evaluated on a confirmed bar.

These alerts fire on the right-shoulder confirmation bar of a pivot pair. They are one-shot per confirmation (the geometry does not continue latching on subsequent bars; another confirmed bullish divergence requires a fresh pivot pair).

The divergence alert honesty beat. The alert fires on the right-shoulder confirmation bar, which is Pivot Len bars after the original pivot bar. Plot On Pivot does not move the alert's bar of record. If the triangle is visually back-shifted on your chart because Plot On Pivot is ON, the alert still fired on the confirmation bar, not on the visually-anchored pivot bar. That distinction is non-negotiable and every mention of divergence on this pack carries it.

Why it matters operationally: a reader who wires the divergence alert to a notification channel sees the alert arrive at the confirmation bar — which is the only bar it can arrive at, because the right-shoulder pivot needed to close before the geometry could be evaluated. In a backtest-style post-hoc review with Plot On Pivot ON, it is easy to look at the back-shifted triangle and feel like the tool reported the pattern earlier than it did. It did not. The alert and the real-time visibility of the geometry are both gated on the same confirmation bar.

The misread this alert attracts. "Bullish divergence alert fired, go long." No. The alert fired because two confirmed price pivot lows in lower-low order paired with two blended-MACD values in higher-low order at matched offsets. That is a description of a completed geometric pattern. It is not a directional forecast. It does not obligate price to reverse, does not obligate volume to confirm anything, and does not obligate the next bar to be consistent with the pattern at all.

Treat the alert as a prompt to check the slot stack underneath. If the stack is already rolling over, the triangle is context that agrees with evidence already forming. If the stack is still confidently trending, the triangle is a geometry note inside a continuation. The triangle is geometry, not a call — and the alert is a notification that the geometry exists, not a recommendation to act on it.

Repeat-fire vs flip, in plain language

Three of the five groups hold state and will fire every confirmed bar where the state holds:

• Per-slot bullish/bearish.

• Blend state.

• All aligned.

Divergence is the exception — its alert is tied to the confirmation of a specific pivot pair and will not re-fire on a subsequent bar without a fresh pair.

If you want a flip-only behavior on the state groups — e.g., "alert me only when blend K crosses blend D, not every bar it remains on one side" — wire flip-detection downstream. TradingView's built-in alert editor supports this through the "Once per bar close" option combined with a flip condition you author separately, or through external triggers on your side. The indicator does not fire flip-only alerts itself because a flip is a reader-specific framing, not a universal one.

Bar-close gating, in detail

Every alert condition AND-gates on confirmedBar = barstate.isconfirmed. In TradingView terms:

• Alerts never fire on tick.

• Alerts never fire on an intra-bar recalculation.

• Alerts fire when the alert's evaluation bar is confirmed.

For per-slot alerts, the evaluation bar is the chart bar (because the slot's state K > D is computed against the chart bar's current view of that slot's K and D). A per-slot alert on a 1m chart with a 60m slot will evaluate every 1m confirmed bar using the current 60m slot read. If that slot's On Bar Close? is ON, the slot reads one HTF bar back and the state is stable within HTF bars. If OFF, the slot reads live and the state can flip within an HTF bar — the alert will fire on the chart bar where the flip occurred.

For blend alerts, same pattern against the blend's own state. For all-aligned alerts, same against every enabled slot's state simultaneously. For divergence alerts, the evaluation is inside a pivot-confirmation block and requires barstate.isconfirmed plus both the price pivot and the matched-offset blended-MACD value to be non-na.

What is deliberately absent

• No threshold-crossing alerts. There is no "blend crossed 70" alert. The reference lines at 30 and 70 are visual bumpers, not thresholds with logic attached. An alert would treat them as thresholds, and the pack refuses to treat them as thresholds.

• No transition-edge alerts as a built-in class. The state alerts hold state. Flip-only detection is left to the reader because the right framing depends on the reader's workflow.

• No per-slot divergence alerts. Divergence is a blend-only feature on STR. If you want per-slot divergence with its own alerts, go up to CTX.

• No alerts on Keltner envelope touches, BBWP threshold crossings, or Donchian channel edges. These features describe the blend's posture — the pack does not frame posture descriptions as alertable events. A reader who wants Keltner-touch alerts or BBWP-threshold alerts has a reasonable request; the pack's position is that an alert on a non-independent structure feature would invite exactly the kind of misread the structure-features-not-independent qualifier is trying to prevent.

Wiring an alert — two examples

Review queue wiring. The per-slot, blend, all-aligned, and divergence alerts route naturally into a review queue — a feed you watch, not a feed that executes. Set "Once per bar close" to keep repeat-fire manageable. Use a description that names the condition in plain English. Route to a notification channel you check deliberately, not one that paginates alongside chat.

Downstream flip detection. If you want a flip-only variant of Blended MACD Is Bullish, add a condition downstream that fires only when the state crosses from bearish to bullish. TradingView's alert system supports compound conditions via external webhook logic or via "Alert on one-bar-change" filters on the receiving side. The indicator itself will keep firing the state alert every confirmed bar; you clip the repeats off downstream.

What the alerts are for

They are state descriptors for traders who want a signal-to-review system rather than a signal-to-execute system. A per-slot bullish alert says "that slot's convergence story is bullish right now, on this confirmed bar." A blend state alert says "the weighted summary of your active slots is bullish." An all-aligned alert says "every enabled slot currently agrees." A divergence alert says "a confirmed divergence geometry just evaluated."

None of them are trade triggers. They commit to state and geometry, nothing more.

For the rest of the misreads — alignment trap in full, divergence-as-entry in full, BBWP-as-price-volatility, structure-features-as-independent-votes — go to Limitations and Trust Boundaries.