Limitations and Trust Boundaries

This page exists because STR's best feature and its most dangerous misread are the same thing: a single bounded pane where multiple structural readings sit inside one coordinate system. That shared frame is what makes the oscillator's shape legible — and it is what makes the pane easy to over-read, because several persuasive elements can agree with each other and still not add up to a decision. The purpose of this page is to be explicit about the four things most often collapsed together in the reader's head: what STR will honestly tell you, what you still have to verify yourself, what can drift quietly as context changes, and which patterns look like confirmation when they are something else.

Keep this page in your working rotation. It is worth returning to after a week of use, after a month, and especially after any session where a trade did not do what the pane seemed to promise — those are the sessions where the most instructive misreads are still fresh enough to examine honestly.

What you can trust

These are statements STR earns. The indicator consistently honors them on any correctly configured chart.

• The pane is bounded. Every plotted line and band — blended fast, blended slow, per-slot lines, Keltner upper, Keltner lower, Keltner basis, Donchian upper, Donchian lower, Donchian basis, BBWP columns — is clamped into the 0 to 100 window at its final step. Nothing will leave that window on a correctly rendered chart. If something appears to, your chart rendering is at fault, not the indicator.

• Divergence triangles only print on confirmed chart-price pivots. The script confirms the price pivot first, then compares the blended fast value at that same pivot offset against the prior sampled value. Triangles that appear during a live bar are reporting a price pivot that has already confirmed, not a speculative one.

• Every alert is gated on a confirmed bar. No developing bar ever triggers an alert. The alerts are honest about when their condition became true.

• The repaint posture is named and exposed. On Bar Close? defaults to true on every slot, which is the safer side of the tradeoff. If you ever flip one to false, you have made that choice knowingly and the MTF and Repainting page tells you what you bought with it.

• Slot timeframes are protected by a hard stop. If any slot's timeframe is below the chart timeframe, the indicator refuses to run and names the specific slot in the runtime error. This is deliberate. Serving a lower-timeframe request from a higher-timeframe chart request cannot be done honestly.

• BBWP will not paint until it has enough history. A short-history chart sees empty columns instead of a noisy percentile rank. This is a trust-building silence, not a defect.

• Enable toggles are real. A disabled slot does not plot, does not alert, and does not contribute to the blend. A disabled extra does not draw and does not fire its alerts.

What you have to verify yourself

STR does not verify these for you. If they matter to your decision, you verify them every time the indicator is applied to a new chart or the context changes.

• That the chart has enough history. BBWP defaults to a 252-bar lookback and needs the full window. Pane elements that depend on historical comparison (BBWP, Donchian, KC smoothing) give their best readings with ample confirmed history on the chart.

• That weights still reflect your intent. Weights do not drift, but your views do. A weight configuration that matched your read of a symbol three months ago may no longer match it. When the pane surprises you, the first place to look is your weights.

• That On Bar Close? is still set the way you think it is. In any per-slot Power User group, the toggle can be flipped and forgotten. If the chart starts behaving noisier than you remember, confirm every active slot's repaint switch.

• That your Ticker: overrides still make sense. A cross-ticker slot can stay configured against a symbol that used to correlate with your main chart but no longer does. The indicator will keep reading it — it will not notice the correlation broke.

• That the sensitivity setting has not been cranked past the useful range. Saturation looks confident and it costs you resolution. Every few sessions, drop ATR Sensitivity back to 1.0 and confirm the pane still looks like you remember. If it suddenly comes alive, you were above the useful band.

• That the blend is still a blend. A weight configuration collapsed to one slot is legal and sometimes desired. It is also easy to forget you did it. Check that at least two slots have meaningful non-zero weights when you expect a blend.

What can drift with context

These are behaviors that are not defects but can change their character as the instrument, timeframe, or market regime changes.

• Sensitivity behavior across instruments. ATR Sensitivity of 1.0 on a liquid major-index chart is not the same feel as 1.0 on a thin small-cap. ATR-normalized distances will scale differently, and the default sensitivity may feel hot or cold on different instruments. Treat 1.0 as a starting point per instrument, not a universal.

• BBWP rank stability across regimes. BBWP ranks recent width against its own history. When the regime shifts — quiet to violent, trending to chop — the rank is ranking one regime against another for a while. During that transition, BBWP can read low when you expected high and vice versa. This is the metric honestly reporting that it is spanning two regimes, not a defect.

• Keltner behavior at the pane edges. When the blended fast spends significant time near 0 or 100, the Keltner envelope does not have symmetric room to express. The band nearest the edge will compress against the clamp. This is correct behavior; treat a clamped band as uninformative in that direction until the oscillator pulls back inside the body of the pane.

• Divergence clarity in chop. Divergences rely on pivots. Choppy markets produce many shallow pivots; Pivot Len: at 20 may resolve too many of them into triangles that do not carry much weight. In chop, a longer Pivot Len: produces fewer but more meaningful divergences. Know which regime you are in.

• Alignment count with long-timeframe slots. A 60-minute or higher slot takes longer to populate on a freshly loaded chart. During the warm-up window, alignment alerts operate on a smaller population than you intended. The alert is still honest; the quorum is smaller.

What not to assume even when the chart is persuasive

These are the specific over-trust patterns that catch experienced readers. Most of them look like confirmation because they combine several elements of the pane agreeing. The mistake is not in the reading; it is in the inference from the reading.

• A confident divergence triangle is not a reversal call. A confirmed divergence is the oscillator and price telling different stories at a pivot. That disagreement can resolve either way. Triangles at extremes are more interesting than triangles in the middle, but none of them are decisions.

• All-slots-bullish is not continuation. Alignment is agreement. Agreement can happen at the start of a move, in the middle of a move, or at the last gasp of a move. Alignment plus BBWP low plus Keltner hug plus the blended fast above 70 looks like a pile of reasons to add; it is often the pattern of an over-extended oscillator about to unwind. Let the pane agree; do not let it decide.

• A Keltner hug is not trend strength. The blended fast sitting on the upper Keltner means the oscillator has been expanding at a rate near its own recent average. That is a statement about the oscillator, not about whether price will keep moving in the direction the oscillator is currently leaning.

• A Donchian break on the oscillator is not a price breakout. The oscillator hitting its recent high is a different event from price hitting its recent high. They can coincide; they often do not.

• A smoothed blend is not a slower market. If you turned on Enable Smoothing, the crosses you see are reported a bar late. A cross that looks crisp on a smoothed chart is a cross that already happened.

• A hidden slot is not an absent slot. Hidden slots still contribute to the blend. Surprise behavior in the blend is often a hidden slot doing exactly what it was told to do, out of sight.

• "Blended" is not always blended. When weights collapse to one slot, the blend is a single-slot reading that kept its old name. Alerts that mention the blend are still reporting that single slot's state.

The trust boundary in plain language

STR is a reading instrument. Every line on the pane, every column, every triangle, and every alert is reporting state — something that was true on a confirmed bar or at a confirmed pivot. None of those statements extend one bar forward. None of them speak about chart price beyond the specific relationships named in the visuals page. None of them evaluate your risk, your position, your account, or the broader context around the instrument you are looking at.

The reader is the integrator. You carry the method and the risk posture. You carry the context about the instrument that STR has no access to — earnings calendar, correlated moves, session behavior, the thousand small pieces of situational awareness that no oscillator can encode. STR is the instrument; you are the operator. That framing matters not as ceremony but because it changes what you do when the pane gets persuasive. A reader who holds STR as a decision system will trust the pane hardest in exactly the situations where the pane is reporting one kind of shape inside a larger context the oscillator does not see.

When the pane feels as if it is about to tell you something definitive, treat the feeling itself as a diagnostic. The pane is not built to be definitive; it is built to support a reading. If you find yourself leaning on it for certainty, that is the signal to widen the frame — check the instrument's context, check whether your own state is adding pressure to the decision, check whether you are tuning to support a bias. None of those checks are in the indicator. They are in the operator. The indicator's job is done when it has reported what it sees honestly; the operator's job starts there.

When the pane will not help you

STR will not do these things, and pretending otherwise is expensive:

• Tell you whether to enter or exit a position.

• Tell you position size.

• Tell you anything about the ask, the bid, the spread, or slippage.

• Evaluate news, earnings, or event risk.

• Make a claim about probability of follow-through.

• Verify itself against a new instrument you have not yet studied.

• Catch the case where your cross-ticker slot is reading a symbol whose correlation broke.

• Correct for saturation you have introduced by tuning.

None of these are failures. The tool is not scoped to any of them. The limitation that matters is not in the code — it is in the reader's expectations. Keep the expectations matched to what the tool actually does and the indicator becomes what it was built to be.