Limitations and Trust Boundaries

This page is for the reader who wants to know where the instrument stops being useful and where it stops being honest. Trust boundaries are not disclaimers; they are part of the instrument. If you cannot name where a tool fails, you are running on belief.

A short framing before the list. Axiom Stoch Osc STR is a lens, not a call. It renders a particular read on multi-timeframe stochastic behaviour with four optional framing features on top of it. It does not know where price is going. It does not know what the next bar will do. It does not know what your risk tolerance is or what your account looks like. Your judgment is not something the tool replaces; it is something the tool feeds.

The rest of this page names the specific failure modes, the specific misreads, and the specific places where you should not trust the output without doing additional work. Each one is either a mechanical fact about the code or a behavioural fact about the reader.

Failure modes that live in the code

The alignment alerts cannot fire when slot 05 is enabled

In the current alignment-counter block, the slot-05 branch increments activeSlotCount six times instead of once. The subsequent equality check between activeSlotCount and bullCount or bearCount can never be satisfied because those counters can only increment once per enabled slot. The result is that neither All Stoch Slots Bullish nor All Stoch Slots Bearish fires on any configuration that has slot 05 enabled.

The manual names this on the alerts page with the full mechanism, on the troubleshooting page as the first-class symptom, and here as a trust boundary. If you enable slot 05 and depend on the alignment alerts, you are building on an alert that will not fire. The workaround is one of:

• Keep slot 05 disabled when alignment alerts matter.

• Build alignment externally from the per-slot bullish and bearish alerts.

• Use the named hidden plot placeholders, {{plot("Active Bullish Count")}} and {{plot("Active Bearish Count")}}, to surface active bullish and bearish counts in alert messages and evaluate alignment in your downstream system.

The manual pack does not fix the code. It tells you what the code does. That is the posture.

Mixed-posture On Bar Close? across slots

Per-slot On Bar Close? is a legitimate feature that replaces Base's single global switch. It also makes a failure mode possible that Base cannot produce: a blend whose members disagree about which bar's close they are reporting.

A slot with On Bar Close? = true returns the previous confirmed value. A slot with On Bar Close? = false returns the live forming value. A blend that mixes both is averaging past certainty with present volatility. Alignment counts treat every slot equally. Structure features read the mixed blend as their input.

Mixed posture is legal. It is not safe in the generic sense. If you cannot finish the sentence "I set slot 03 live and slots 01 and 02 confirmed because ____," your safer position is uniform true across every enabled slot. The mtf-and-repainting page walks through the mechanics.

Slot TF below chart TF halts the script

The script refuses to run when any enabled slot's timeframe is lower than the chart's timeframe. It throws a runtime error naming the exact slot. This is a design constraint, not a bug. The workaround is to move the chart down to a timeframe the slot can legally read, raise the slot to the chart timeframe or higher, leave the slot empty to inherit chart TF, or disable the slot. The troubleshooting page carries the full row.

Cross-ticker slots inherit their symbol's session

A slot with a non-empty Optional Ticker reads its source on that other symbol. The other symbol's session governs when the slot has data. An equities symbol slot on a 24-hour futures chart will show gaps or flat values outside the equities session. The blend inherits those gaps. The divergence module still compares chart-price pivots against the blended K, which can produce noisy divergences around the session boundary.

This is not a bug — it is what cross-symbol work requires. The trust boundary is: if you enable a cross-ticker slot, verify its session behaviour against your chart's session before trusting any alert that combines them.

Misreads that live in the reader

The four structure features are not four independent confirmations

Divergence, Keltner, BBWP, and Donchian all read the blended K line after master smoothing and the final clamp. They are four lenses on the same series. When a divergence triangle confirms and the blend is near a Keltner band and the BBWP has just moved and the blend is pressing a Donchian edge, that is one line agreeing with itself.

Reading the four features as four independent votes builds false confidence in exactly the moments where the pane is least informative. The correct read is the opposite: when several features move together, ask what changed in the blend — a slot weight shift, a master smoothing pass, a cross-ticker slot's session boundary — rather than treating the co-motion as confirmation.

Divergence is a descriptor, not a call

A confirmed bullish divergence is a particular geometric pairing between two chart-price pivot lows and two blended-K values at the corresponding offsets. That is a fact about the pivot that just completed. It is not a fact about the next bar.

A reader who enters on the confirmation bar because a lime triangle just printed is trading on a lagged descriptor. The triangle is Pivot Len bars to the right of the pivot it describes — by construction. Price has had Pivot Len bars to do whatever it was going to do by the time the triangle confirms. The divergence is a question worth asking: "chart price made a lower low while the blended K did not." What you do with that question is your work, not the indicator's.

Plot On Pivot is an honest drawing, not an honest timing

Plot On Pivot on back-shifts the triangle onto the pivot bar. The triangle was not visible at that bar when the bar was printing. The alert still fires at the confirmation bar. Treating the back-shifted marker as evidence you could have acted on at that earlier bar is a decision made on evidence that did not exist yet.

Use Plot On Pivot for readable charts, not for retroactive post-hoc reasoning about what you could have done. The visual is honest about where the geometry sits; it is not honest about when you could have seen it.

80 on the blend is not 80 on a classical fast stochastic

Single-symbol single-timeframe stochastic habits — watch 80 and 20 for reversal, trade K-D crosses inside the extremes, read "hooks" from the extremes — assume a stochastic running on one symbol at one timeframe with one lookback. The blend here is a weighted mean of up to five slots, each potentially on a different timeframe, a different symbol, or a different source.

The blend reaching 80 says the weighted mean of the slots' K lines is in the upper reference zone. The next move is not implied by any classical-stochastic folklore applied to that number. Blend-level reversal-at-80 trades are a transfer of habit from a different instrument, not a validated use of this one.

BBWP is oscillator width, not price width

BBWP columns are a Bollinger band width percentile computed on the blended K line. A short blue column means the blend's band width has been compressed against its own history. It does not mean price has been quiet. A tall aqua column means the blend's band width has expanded above the configured threshold. It does not mean price has been volatile.

Sometimes the two correlate; often they do not. If you treat BBWP as price volatility, you will draw wrong conclusions in exactly the regimes where oscillator and price volatility diverge. Treat it as what it is: a statement about the oscillator's own recent shape.

Keltner touch on the oscillator is not price overbought or oversold

The Keltner envelope wraps an MA basis of the blended K with a bar-to-bar range estimator on the blend itself. A touch on the upper band means the blend has stretched against its own recent range. That is a statement about the oscillator's stretch. It is not a statement about price being overbought. It is not a reversal signal.

Overbought and oversold in the classical sense live at the 80 and 20 reference guides. Keltner is a different question on a different scale. Do not substitute one for the other.

Hide Plot is not a kill switch

Hide Plot removes the slot's visible K line from the pane. The slot still runs. The slot still contributes to the blend by weight. The slot still drives its per-slot alerts. A reader who hides slot 03's plot and later forgets about it will be confused when the blend moves for reasons the visible slots cannot explain. Open Settings to see the real state; do not rely on the pane to tell you which slots are active.

Weight zero is not a kill switch

Setting a slot's Blended Weight to 0 silences its contribution to the blend. The slot still runs. The slot still plots if Hide Plot is off. The slot still fires per-slot alerts. Readers who zero a slot's weight to "turn it off" will still receive per-slot bullish and bearish alerts from a slot they believed was muted.

For a real kill switch, use Enable Stoch 0X = false.

Master smoothing delays everything

Enabling master smoothing on the blend calms the K and D plots. It also delays every consumer: the blend K/D cross fires later, the divergence module reads a smoother input so pivot comparisons are slower, Keltner and Donchian and BBWP all read the smoothed series. Per-slot alerts are unaffected because they evaluate the slot's K vs D, not the smoothed blend.

If you enabled master smoothing because the blend looked choppy and then treat the calmer alerts as more reliable, you traded freshness for quiet and missed the tradeoff. When the chop was evidence, the calmer plot is worse information.

Changing Source does not change the stochastic range

ta.stoch(src, high, low, n) takes a source argument and a separate high and low argument. The script always passes the bar's own high and low to the range denominator regardless of the source. Changing Source from close to hl2, hlc3, or ohlc4 only moves the numerator — where inside the bar's range the current value sits.

Readers expecting "Source = HLC3 means the stochastic range rescales" will misread the knob. On narrow-range bars the K line will barely change when you change Source because the bar's high-low range did not change.

Why this page is this long

The limits above are not disclaimers. They are the list of failure modes the pack has decided to make visible on purpose, in the same voice it uses for features. The reason that choice is worth naming is that it has a cost. A page that reads "here is everything that can go wrong with this tool" is a harder page to sell from than one that keeps the failure modes in an appendix. Most vendor documentation does the appendix version. This pack does not.

The working logic behind the decision:

• An undocumented failure mode is still a failure mode. The alignment-alert behaviour with slot 05, the mixed-posture repaint risk, the four-features-share-an-input fact, and the stochastic range asymmetry all exist whether or not this pack writes them down. They cost the same whether they are surfaced or buried. The only thing publication affects is whether you find out before you commit real money to a conclusion that relied on the hidden edge.

• A reader who cannot name the tool's failure modes is running on belief. The entire point of the pack is that you read the pane yourself, which requires that the pane not be misleading you in ways you do not know about. The page above exists to remove specific forms of not-knowing.

• Recommendations would be worse than silence. You will not find a "best settings" table here, or a "works especially well on EURUSD" claim, or a recommended configuration for swing versus scalp. Those would be the teaching-shaped versions of marketing. The tool is adaptable for the reader's work; the tuning choices are the reader's work. A packaged recommendation would take that work away, which is also taking away most of the value.

The practical consequence for your reading is that the pages are denser than a typical vendor pack and the honesty is front-loaded instead of back-loaded. If that density feels unusual, it is because the standard shape of documentation has been letting you skim past the parts you need most. This pack is written for the reader who would rather spend an hour studying the failure modes once than spend a week discovering them unaided.

Where the instrument stops and you start

The instrument can do:

• Compute a weighted mean of up to five stochastic slots across up to five timeframes or symbols.

• Keep the pane in a stable 0..100 space with hard clamps at every stage.

• Provide four optional framing features on the blend: divergence, Keltner envelope, BBWP columns, Donchian steplines.

• Gate every alert behind confirmed bars so forming-bar noise does not fire alerts you cannot unwire.

• Expose sixteen alert conditions for per-slot, blend, alignment (with the slot-05 caveat), and divergence states.

• Let you choose the MA family for every smoothing step and for every structure feature's basis, with Power User blocks for families whose behaviour is not captured by length alone.

The instrument cannot do:

• Know what price does next.

• Tell you whether a divergence will be followed by a reversal, a continuation, or a chop.

• Replace your read of price action, volume, session context, correlated symbols, or the specific instrument you are trading.

• Guarantee that a configuration works on every symbol, every timeframe, every regime.

• Warn you when your mixed-posture configuration is producing a blend that mixes confirmed and live pieces — you have to check On Bar Close? on every slot yourself.

• Distinguish "four structure features agreeing" from "one series seen through four lenses" — the pane shows both identically; you have to hold the difference.

Your judgment starts where the instrument stops. That is not a bug — that is the whole proposition.

Where to go next: Alerts for the alignment disclosure in its wiring context, MTF and repainting for the mixed-posture failure mode in depth, Workflows for the anti-patterns named in context of real read sequences, For the geeks for the mechanical picture that makes the trust boundaries concrete.