Limitations and Trust Boundaries

Every teaching page in this pack gives you the shape of the tool. This page gives you the honest edges. Read it in order on the first pass. The first section is the one that costs the most when it is ignored, so it sits at the top.

A note on posture before you read. None of these are disclaimers. They are pieces of the instruction. A reader who can name these limits can read the pane through noise and pressure; a reader who cannot will misread the pane under exactly the conditions where correct reading matters most.

1. Alignment is not breadth

This is the section that matters most. If you remember one thing from the pack, this is the thing.

The All MACD Slots Bullish and All MACD Slots Bearish alerts describe a state: every enabled slot with a non-na value sits on the same side of its own signal. They do not describe independent evidence. The difference between "state across my configuration" and "three independent observations agreeing" is where almost all of the damage lives — not because the alert is dishonest (it reports what the configuration is currently saying) but because the name "All Slots Bullish" is easy to translate into "every timeframe independently concluded bullish," and that translation is wrong in most real configurations readers build.

If your three slots share source, MA family, and length pair, and differ only in timeframe, you have a single MACD read sampled at three cadences — not three observations that can independently agree. On trending instruments, that kind of alignment holds for most of any trending session. You will see the alignment alerts fire reliably. The reliability is a feature of the configuration, not evidence for the read. The alignment is behaving less like a consensus detector and more like a "the trend is persisting" detector. Those are different things.

In Base, the cost per slot is higher than in CTX. With three slots carrying a third of the blend each, a single mis-differentiated slot is pulling a third of the blend and a third of the alignment signal. On the ten-slot CTX context surface, a single duplicated slot is pulling a tenth. That does not make sloppy CTX configuration harmless, but it does mean Base punishes duplicated intent faster because each slot carries more of the read.

The cure is differentiation, not suspicion. Three slots that genuinely represent three different stories — different length pairs, possibly different MA families, possibly different sources — produce an alignment alert that fires when three different readings actually agree. That is a meaningfully stronger signal about agreement than the same alert on three copies of the same read. Workflows shows the shape in Scenario 1; the contrast with Scenario 3 is the shape you are trying to stay out of.

And one more wrinkle that catches readers who use the observer pattern: alignment counts enabled slots regardless of weight. A slot with weight 0 that is still enabled is still a vote in the alignment tally. If you want a slot out of the alignment conversation entirely, disable the slot.

2. Boundary pinning is not intensity

A slot reading pinned against 100 for several bars is not telling you the convergence gap is more stretched than a reading at 80 would be. It is telling you the sensitivity-and-volatility pairing on that slot has pushed the mapping into its flat region, where different raw values compress to approximately the same number. Pinning is a saturation artifact, not an amplification of the read.

Two signs that pinning is the issue rather than a real event:

• Only one slot is pinned while the others breathe. If the pane is reading an instrument-level intensity, you would expect it to show up across multiple slots, not just the one whose ATR and sensitivity happened to combine this way.

• The raw price action on the chart does not look outsized. If the chart bars are average for the instrument's session and a slot is pinned, the reading is a projection artifact, not a market statement.

The fix, when you are ready for it, is a small ATR Length increase, a small ATR Sensitivity decrease, or both — and then watching for a session before deciding whether the new pairing is better. The wrong fix is cranking sensitivity higher to "make the pane more responsive" when you see pinning; that makes the flat region wider and the pinning more frequent. Settings walks the sensitivity tradeoff in the tripwires block.

3. A quiet blend on differentiated slots is not a calm market

The blend is a weighted average. If your three slots disagree — one bullish, one near 50, one bearish — the weighted average will sit near 50, and the blend-based alerts will not fire, and the blended line will look calm. The calm is a calm on the blend, not on the market. The slots themselves are carrying a disagreement story that the blend averages over.

This is why Visuals and Logic insists on reading the slot lines before the blend. A quiet blend on three differentiated, actively disagreeing slots is a signal that three different timeframes are telling three different stories — which is often more interesting than a clean three-way agreement, not less.

4. The histogram tells speed, not direction alone

Reading the four-state histogram as a simple green-up, red-down surface throws away half of what it encodes. The histogram is a two-dimensional reading — side of 50 crossed with rising/falling — compressed into one color glance. A reader who scans "green or red" and not "bright or faded" is trading on half the information.

The most common missed state is faded green. It reads as "still green, still bullish, no change needed." What it is telling you is that the convergence story is bullish and losing thrust — the thrust is decelerating, which is often the earliest visible sign that a regime is about to shift. Dismissing faded green as "still up" is how readers get caught flat-footed on reversals that the histogram was already warning about.

Same asymmetry applies on the bearish side. Faded red is bearish and losing thrust. It is not "still down, keep short"; it is "the bearish thrust is decelerating, stay alert for a turn."

5. A weight-zero slot is not a silent slot

Setting a slot's weight to 0 excludes it from the blend. It does not disable it. The slot still computes, still plots (unless you also hide the plot), and — this is the one that bites people — still fires its per-slot alerts and still counts toward the alignment alerts.

If you are using the observer pattern on purpose, keep it documented. Template names, chart annotations, or a short comment in your workflow all work. Undocumented zero-weight observers are the kind of thing a reader discovers a week later when a pane moves in a way the visible, voting slots cannot explain.

If you want a slot out of everything — blend, alignment, alerts — disable the slot. Weight 0 is a specific behavior, not a way to turn a slot off.

6. Master smoothing is lag, not truth

The optional post-blend smoothing pass makes the blend calmer and can change blend-alert timing. It does not make the blend more correct. Chop on the unsmoothed blend is evidence about the slot evidence; hiding the chop with a smoothing pass does not repair the evidence, it just hides it.

Two specific consequences to hold onto.

• A blend-bull alert on the smoothed blend is not the same alert timing as the unsmoothed blend. It often arrives later, and short-lived raw blend states may not survive the smoothing pass. Reading the smoothed-blend alert as "clean early confirmation" overstates what smoothing can do.

• Per-slot alerts and per-slot lines are not affected by master smoothing. So a reader who enables master smoothing "for cleaner reads" gets a cleaner blend and unchanged slot evidence — which means the slot alerts may be firing earlier than the smoothed blend alerts, and the reader has to be clear about which they are acting on.

If you enable master smoothing, do it because you want a slower regime view. Do not do it because you want noisy alerts to stop. The fix for noisy alerts is a longer length pair on the underlying slot, not a smoother wrapper on the blend.

7. Chart-timeframe collapse is silent

This is covered in full on MTF and Repainting. Restated briefly here because it belongs in the limits conversation:

When the chart sits at or above the slowest configured slot timeframe, the slots compress toward the chart's own cadence, and the pane stops being a multi-timeframe pane. The tool does not throw an error — it is honoring the slot-at-or-above-chart rule — but the configuration has lost its teaching purpose. Three slot lines move together. The four-state histogram stops carrying cross-timeframe information. Alignment alerts fire more often and mean less.

Noticing it requires checking the chart timeframe against the slowest slot. The fix is either a lower chart timeframe or a slot ladder that has genuine headroom above the chart.

8. Short-slow noise is a cost, not a signal

A slot with very short length pairs (for example, 3/7/2) will flip direction frequently on any instrument. That is not the pane being noisy; that is a very short MA difference being the unsmoothed quantity it always is. A reader who sets short length pairs to "make the pane more sensitive" and then treats every slot-line color flip as evidence is reading a noisy series as if it were a clean one.

If short pairs are the right choice for a workflow, the corresponding cost is a higher false-turn rate on that slot's line and on its per-slot alerts. You either accept the cost or you move the length pair up. There is no third option where short pairs are simultaneously reactive and quiet.

9. On Bar Close? = off is a choice, not a speed upgrade

Off is not "the faster mode." It is "the mode that exposes live slot-bar evidence at the cost of that evidence being capable of revising before the slot's current bar closes." The update speed you gained and the confirmation stability you lost are two sides of one coin; taking the first without accounting for the second is how readers end up acting on readings that later unwind.

Concretely: on a 1-minute chart with the default 60-minute slot 03, off mode lets slot 03 move during the live 60-minute bar. That motion can go either direction, possibly several times, before the hour closes. A blend-bull alert that fires at 10:18 on the smoothed, off-mode blend can be perfectly correct at 10:18 and still be wrong in its implication by 11:00, because the 60-minute evidence behind it has revised. The alert fired on a confirmed 1-minute chart bar using slot evidence that was still forming; there is nothing dishonest about the fire, but "confirmed on the chart bar" is not the same as "confirmed on the slot's own bar."

MTF and Repainting walks the direct verification for this. Run it once. It takes five minutes and it is the fastest way to feel the tradeoff rather than merely read about it.

10. Defaults are scaffolding

The default three-slot stack (5 / 15 / 60, 12/26/9 EMA on each, equal weights) is what loads on install. It is not a recommended production configuration for your instrument. It is a useful starting point for learning the pane's behavior. A reader who runs the defaults for a week and then declares them working is actually declaring that the pane-on-defaults produced a workflow that felt okay; that is not the same as the defaults being right for that instrument, that session length, or that risk profile.

Tune them the way the pack asks you to: one knob at a time, with a question you can state in words, held across a session, and rejected if the session says no. Do not optimize to a single good-looking week.

11. Non-repainting without qualifier is a claim this pack refuses to make

Every multi-timeframe tool has a repaint question. This one puts the answer on a switch and names both sides of the switch. The honest statement is: when On Bar Close? is on, the slots return confirmed previous-bar values; when it is off, they return live intra-bar values that can revise. Anything that reads as "non-repainting" without that qualifier is leaving out the relevant sentence.

12. This pane does not produce entries

Ten alert conditions. None of them are entries. Alignment across three differentiated slots is not an entry. A blend-bull fire is not an entry. A bright-green histogram column is not an entry. Those are readings that could be inputs into a decision you make under whatever rules you already trust; the indicator does not tell you to click a button, and treating any alert here as a trigger is a misuse of the alert and of the pane.

The line to walk is this: the pane is instrumentation. It shows you the convergence/divergence story across three timeframes in a bounded frame. What that story means for your position sizing, your risk rules, your entry checklist, your time of day, your instrument, and your session — those are decisions only you can make, and only on top of a reading that you can defend. If that separation feels like an omission, the rest of the Axiom tool family (strategy builder, structure tools, other oscillators) is where the conversation about decisions lives. Reading this pane correctly is a prerequisite for that conversation, not a substitute for it.

Where to go next

• The named workflows that make the alignment conversation concrete — Workflows.

• The knobs that push pinning and sensitivity drift in either direction — Settings.

• The conceptual shape of the ATR-sigmoid mapping, without reproducing it — For the Geeks.

• When a symptom on the pane suggests a setup error rather than a misread — Troubleshooting.