Limitations and Trust Boundaries

Every tool asks to be trusted. The ones worth trusting tell you, up front and in plain language, what they can and cannot do — so the trust you extend is a decision you made rather than a default you inherited. This p...

Written By Axiom Admin

Last updated 22 days ago

Limitations & Trust Boundaries

If you skim one page in this pack, skim this one. The content below is not a caveat strip bolted onto a sales page. It is the set of honest boundaries the rest of the pack sits inside of, and a reader who understands these boundaries can use the pane with real confidence inside them. A reader who has not seen them will eventually discover one by trading against the wrong assumption, and the discovery will cost them more than the ten minutes it takes to read this page.

The estimate boundary — the one every page defers to

CVD in this indicator is an OHLCV-based delta estimate. It is not footprint-grade bid/ask delta. The distinction matters and is the most important single thing for a reader to internalize before relying on the pane.

What "OHLCV estimate" means

The estimator reads, for each bar at the slot's lower-timeframe precision, the open, high, low, close, and volume. From those five numbers it builds a per-bar intent score that attempts to infer whether the bar was more likely the product of buying or of selling pressure. It does this using three components — body direction, close location within the bar's range, and wick asymmetry. Wick Weight shapes how much wick evidence enters the intent score; Pressure Sensitivity shapes how decisively that score is classified and carried through ambiguous bars. The resulting signed volumes accumulate across the slot's window. The accumulation is normalized against its window's high-low range into the 0-to-100 pane.

This is a defensible estimator. It captures information that a naive "close greater than open means buying" reading misses — balanced-body bars with rejection wicks, decisive closes inside a narrow range, and the rhythm of indecision across a slow stretch. It is the reason this pane is more informative than a cumulative-close-minus-open toy on the same data. But it is still an estimate.

What "footprint-grade bid/ask delta" means, by contrast

A footprint chart receives every tick of trade data with the side marker — which side of the spread the trade hit, which was aggressive — and sums that into the bar's actual buying minus selling volume. When a footprint says "+120,000 delta on this bar," it is reporting the observed net aggressor volume. There is no inference. It is a measurement.

This indicator has no access to tick-level aggressor data. No Pine-only indicator does. The OHLCV estimate is the best defensible read a CVD tool built on OHLCV can produce. It is not footprint. No setting inside this indicator turns it into footprint.

What this means for how you use the pane

Use the pane for directional pressure reads at the estimator's native resolution. "Is the 60-minute stack leaning up or down? Is the daily regime bullish or bearish? Is short-term pressure disagreeing with longer-term pressure?" — all fair.
Do not use the pane to resolve single-bar net-volume questions. "Was this specific bar +50,000 or -50,000 in aggressor volume?" is not a question the OHLCV estimate can answer honestly, and no knob inside this indicator improves the resolution to the point that it can.
On trades where net-volume precision is load-bearing, pair the pane with a footprint tool. The two are complementary, not substitutable.

Every time the word "CVD" appears in the pack in a trust-sensitive context, this boundary is the one behind the sentence. It does not get restated on every page by accident.

The intrabar-history limit

The estimator walks the slot's lower-timeframe precision window inside each slot-TF bar to build its read. That walk requires intrabar history to be available from TradingView. Two limits live on that availability.

TradingView plan and symbol limits

TradingView's paid plans carry more intrabar history than its free plan, and some instruments have deeper history than others. On the free plan, or on a thin symbol, intrabar history runs out sooner as you scroll back through time.

What the pane does when intrabar history runs out

The slot does not error or blank. It falls back to a single-bar OHLCV delta using only the slot-TF bar's own O, H, L, C, and V. The classifier still runs, but it is working with five numbers per slot-TF bar instead of several dozen intrabar samples per slot-TF bar. The output is necessarily cruder.

The pane does not visually announce the fallback. There is no marker, no color change, no indicator on the line that says "everything to the left of here is fallback." A reader studying old bars has to remember that some of what they see on deep historical ranges may be fallback output, and interpret accordingly.

What to do about it

For live and recent-history decisions, fallback is less likely, not impossible. Liquid symbols on recent bars usually have enough intrabar depth for the full-precision estimator. Thin symbols, unusual sessions, or plan limits can still make fallback relevant sooner than you expect.
For backtests across months or years, take the fallback seriously. The historical stretch your backtest is reading may include fallback output. Any "edge" a backtest finds in bars deep enough to be in the fallback is built on a cruder estimator than the live one you would use going forward. Spot-check the backtest's behavior against the pane on recent bars before trusting the result.
To look for your personal fallback edge, scroll back deliberately. Pick a slot and scroll back slowly. The line may become less responsive, smoother, or less informed when intrabar history runs out, but the pane does not give you an exact marker. Treat any deep-history read as lower-fidelity unless you have checked it against recent full-precision behavior.

The runtime guardrails

The indicator refuses to render three configurations that cannot be read honestly. These are listed in detail on the MTF & Repainting page; summarized here because they are real limits on what configurations the pane accepts.

A slot's TimeFrame: must be greater than or equal to the chart's timeframe. Otherwise the pane refuses and the status row names the slot. A 1-minute slot on a 5-minute chart cannot be represented at chart resolution without losing the information that justified using a 1-minute slot in the first place.
A slot's Lower TF Precision: must be strictly lower than its TimeFrame:. A precision equal to or coarser than the slot timeframe has nothing to decompose.
A slot's Window: must be greater than or equal to its TimeFrame:. A window finer than the slot itself has no honest interpretation.

These are not warnings. The pane clears on any of them and displays the runtime error in TradingView's status row.

The guardrails are honest limits, not flaws. Their existence is the cost of an estimator that refuses to silently average over configurations that cannot be rendered truthfully.

The repaint posture under OFF

The MTF & Repainting page is the home for this topic; a one-paragraph version belongs here because it is a trust limit and a reader is entitled to see it listed among the honest limits.

When On Bar Close? is set to OFF on a slot, that slot's value drifts during the formation of its own higher-timeframe bar. What you see at any moment before the bar closes is provisional. The value will be overwritten, potentially several times, before the bar finalizes. The state read from the pane during that drift period is not a confirmed state.

This is not a bug. It is a named, documented tradeoff behind a switch that defaults to the honest posture. But it is a real limit on what an OFF slot commits to before its bar closes, and it belongs on the limits list.

The single-market, single-venue assumption

Each slot reads one symbol's OHLCV. When the symbol is the chart's symbol, you are reading the venue and market TradingView sends. When the symbol is an Optional Ticker:, you are reading whatever venue and market that symbol's data feed is on.

What this means in practice:

A single slot does not aggregate across venues. CVD 01 reading BTCUSDT on one exchange is reading only that venue's view of BTC. If BTC's flow is fragmenting across many venues, a single-venue slot will miss the aggregate story. Some readers counter this by running several optional-ticker slots each pointed at a different venue of the same asset, weighted and blended — a real use case documented in Workflows.
Session behavior follows the symbol's calendar. A slot on SPY with a D window resets at the US equity cash open. A slot on BTCUSDT with a D window resets at UTC midnight. If those two slots are on the same pane in Session mode, their dashed reset markers will land at different wall-clock edges. Rolling mode avoids the mismatch.

What the pane does not do

A list written deliberately as "what is not here," because the absence is the information.

No trade signals. No buy arrow, no sell arrow, no "breakout." Slot color, alignment, blend crossings, and alerts are state descriptors. Converting them into trade triggers mechanically is an anti-pattern documented in Workflows.
No direction prediction. The pane does not know where price is going. It is reading participation context, not forecasting price.
No correlation or cross-asset statistical analysis. An optional-ticker slot lets you see another asset's pressure read alongside your chart's. It does not compute correlation coefficients, beta relationships, or lead-lag inferences between the two.
No divergence detection. The pane does not inspect price action. It cannot say "CVD diverged from price here."
No order-flow events. No absorption markers, no spoofing detection, no iceberg inference. Those are tick-resolution, bid/ask-level questions; the OHLCV estimator is the wrong tool for them.
No regime classifier. The pane does not label states "trending," "choppy," "ranging." That read is yours to make from the slots.
No alert on guide crossings. The 80 and 20 reference levels do not fire alerts. They are cosmetic. See Alerts for what does and does not get an alert condition.

If your workflow depends on any of the above, this indicator is not the right tool for that part of it. It may still be right for the part it is built for, in which case pair it with a tool that covers the rest.

What bounded normalization does and does not guarantee

The 0-to-100 pane is a hard guarantee: no slot line, and no blend line, will ever print outside that range. That property is mechanical, not aspirational.

What bounded normalization does not guarantee:

It does not guarantee the 0-to-100 scale is equally informative across slots. A slot whose window is very quiet may compress most of its activity into the 40-60 band; another slot with a volatile window may spread its activity across the full range. Both sit in the same pane, so a reader has to compare positions as "this slot near the top of its window" rather than "this slot is the highest."
It does not guarantee that an 80 reading is "extreme" in any market-objective sense. 80 means "near the high of what this slot has observed in its current window." On a slow session with no real activity, even 80 can correspond to trivial participation. The pane is bounded, but quiet windows can still print extremes. Cross-check against the slot spread and the blend.
It does not guarantee portability across fundamentally different kinds of instruments. You can run the pane on SPY and on BTCUSDT and compare shapes — that is the point of the bounded scale. But an illiquid penny stock and a major index future may produce pane shapes that are mathematically bounded but operationally incomparable. The bounded scale keeps the visuals legible; it does not make the underlying tape commensurable.

What this pane is deliberately not optimized for

Two honest non-fits worth naming.

Very short, one-trade decision windows. The pane is built for reading context across one to three meaningful timeframes. A reader trying to use it for sub-minute decision-making will find the default defaults are too slow and the fast defaults are too noisy. A simpler, single-timeframe trim of this family (without the CTX bench) will teach the core read faster and with less overhead; come back to CTX when the simpler tool feels too narrow.
Systematic rule-based execution without human review. Every alert in the pack is a state descriptor; none is wired to be a trigger. A systematic loop that fires on CVD 01 Is Bullish without any other gating will fire on every confirmed bar CVD 01 is bullish, which is not "when the state flipped" — it is "continuously while the state holds." The pane can feed a systematic process that includes its own debouncing and gating; it is not intended to be the last hop before execution.

The honest-limits posture, spelled out

The indicator is an instrument for looking at several CVD reads at once, on the same bounded scale, inside a framework that surfaces the estimation choices rather than hiding them. Its delta is an OHLCV-based estimate, not a footprint measurement. Its intrabar precision depends on what TradingView is willing to serve for the symbol and plan the reader is on, and the fallback behavior is silent on the pane even though it is documented here. Its repaint posture is a per-slot switch named On Bar Close?; the default is the honest one, the alternative is legitimate under documented conditions, and a reader who flips the alternative without reading MTF & Repainting is choosing a tradeoff they have not seen.

The guardrails are real. A slot configured to read a timeframe lower than the chart's, or a precision coarser than its own timeframe, or a window finer than itself, will refuse to render. That refusal is part of how the pane keeps the reader inside the configurations it can represent truthfully.

The limits listed on this page are not a disclaimer. They are the boundary of what the tool commits to being. Inside that boundary, it is a careful and legible instrument. Outside it, it is the wrong tool for the question being asked, and the honest response is to reach for a different one.

Where to go next

For the repaint switch and the MTF posture in detail, MTF & Repainting.
For the mental model behind the estimator, For the Geeks.
For workflows that use this tool inside its limits, Workflows.
For what the pane shows and what deserves attention versus what should not be over-read, Visuals & Logic.
For symptom-driven diagnosis when something on the pane surprises you, Troubleshooting.