Limitations and Trust Boundaries

This is the honest inventory page. What the pane can show you, what it cannot show you, and where the estimator's reach actually stops. Every other page in the pack teaches how to use the indicator; this one teaches how much to trust it, by being explicit about its limits.

A word on framing. Candor on this page is not a legal disclaimer, a caveats strip, or a hedge. It is part of how the instrument earns the right to be read. A tool that does not name its own boundaries is a tool that asks you to inherit its confidence without checking it. This page is here so you do not have to do that.

If that sounds self-congratulatory, it is not meant to. Plenty of tools hide their boundaries because naming them is uncomfortable. That is the point. Naming them is the work; it is how a reader decides whether the tool belongs in a decision that will move real money. Skip to the section that matches the question you are sitting with — the page is structured so you can.

The single biggest thing to internalize

This pane plots an estimate, not a measurement. Every CVD value on this indicator — per slot, blended, raw, or smoothed — is the output of a participation model applied to OHLCV data at a lower timeframe. It is a defensible estimate, and For the Geeks walks through the model at the altitude that lets a serious reader decide how much weight to place on it. But it is an estimate. No knob inside this tool turns it into tick-resolution bid/ask delta. No configuration choice trades away the estimate framing. If your process needs footprint-grade data, this is not the right tool, and no page in this pack will try to convince you otherwise.

You will see the word "estimate" in several places in this pack. That is deliberate. Trust-sensitive sentences carry the framing because a reader under stress skims, and the skimmable version has to be honest.

What the pane can show

• Participation-weighted pressure, relative to the window each slot observes. A slot line near the upper band says the cumulative participation-weighted delta is near the upper end of what the slot's window has seen. A slot line near the lower band says the opposite. The statement is honest about the range; it is not a statement about the instrument's absolute pressure.

• Changes in pressure state on a per-slot basis. Slot color flips when the slot's CVD crosses its signal. The color change is a statement that the pressure state, as the estimator sees it, has changed.

• Agreement or disagreement across timeframes. Three slots on three timeframes either agree or they do not. The stack makes that visible in a way a single CVD line cannot.

• A weighted summary of the stack. The blend is the weighted average of the active slots. It is a useful shortcut when the slots agree and a compressed headline when they disagree.

• Window rollovers in Session mode. The dashed vertical markers show you where a slot's cumulative reset landed.

What the pane cannot show

• Who is on the other side of the tape. A slot in the upper band is describing net participation, not the presence or absence of counter-pressure. Large resting liquidity on the opposite side can absorb the participation the pane is reporting without the pane knowing.

• Whether the pressure is aggressive or passive. The estimator reads bar composition; it cannot distinguish an aggressive taker from a passive joiner. That distinction lives in orderbook data, not in OHLCV.

• The orderbook. The pane is blind to resting orders, stop clusters, iceberg orders, and the full book shape. If your process depends on those, this pane is supplementary to a different tool, not a substitute for it.

• Tick-level delta. The estimate walks intrabars at the Lower TF Precision you configured, which is lower-timeframe OHLCV, not ticks. A one-minute intrabar view is a reasonable intraday compromise; it is not the tick tape.

• Future price. The pane describes pressure as the estimator sees it now. It does not forecast. It will sometimes move coincidentally with price in a way that feels predictive; that is coincidence, not prediction, and treating it as prediction is the category of mistake this pack is built to help you avoid.

• "Hidden" participation below the intrabar resolution. The estimator sees what its Lower TF bars see. Sub-intrabar activity is not part of the read.

The estimate boundary, in practice

Whenever a sentence on any page of this pack reads as if the CVD value is a direct measurement ("delta is at 74," "pressure is running strongly bullish"), mentally insert "estimated" in front of it. The pack tries to do that in the writing where it matters most, but the reader's own habit is the most reliable backstop.

Three places the boundary is most relevant:

• Close reads on a single bar. A bar that looks decisive on price can produce a small or ambiguous delta estimate because the estimator reads body, close location, and wicks together rather than just close direction. That is the estimator doing what it is designed to do, not an error.

• Extended stretches of pinning. A slot pinned at 0 or 100 for a long time is a pattern, not a certainty. It can be genuine sustained pressure, or it can be a quiet window being amplified by normalization. The estimator cannot tell you which; your read of the instrument has to.

• Cross-instrument comparisons. The estimator runs bar-by-bar on each instrument's own data. Two instruments' pane reads are not automatically comparable. Similar numbers on two different symbols can represent very different absolute participation profiles. The base trim does not support cross-symbol slots; that feature lives in the CTX trim.

The silent intrabar fallback

This one is worth its own section because a reader cannot see it happen.

When you configure a Lower TF Precision for a slot, the estimator walks that lower timeframe's intrabars inside each slot bar and sums participation-weighted volume. That is the high-precision mode.

When intrabar data is unavailable — TradingView's intrabar history window has a finite depth, and certain feeds do not carry intrabar data for certain symbols — the slot falls back to estimating delta from the slot bar's own OHLCV. The fallback is a cruder read. It still follows the participation model; it just has one bar to work with instead of many, and the model has less material to separate body, close location, and wick from each other.

The line does not announce the fallback. There is no warning, no different color, no marker. You notice it by the line looking blockier on older bars, or by the slot's character feeling less resolved than it did on recent bars. A reader who scrolls backward in history and wonders why the line "used to look sharper" is probably looking at the fallback.

Three practical consequences:

• On recent bars, you almost always have intrabars. TradingView serves them on the active window.

• On historical bars beyond the intrabar window, you probably do not. The line will be cruder there, not because the market was cruder, but because the estimator has less to work with.

• On sparse-data instruments, the fallback can be the default mode. Illiquid symbols or exchange feeds that do not provide intrabar data will run the fallback continuously. The pane will still work; the estimate is just coarser.

MTF and Repainting carries more of the timing detail, including the verification moves you can use to feel the fallback without having to trust that it is happening.

Runtime errors are guardrails, not bugs

The script throws three runtime errors by design. Each of them names the slot in the error text.

• Slot timeframe lower than chart timeframe. A slot cannot produce a resolution below the chart's own bar. The error reads along the lines of "CVD 01 timeframe cannot be lower than the chart timeframe."

• Lower TF precision greater than or equal to slot timeframe. The intrabar walk requires a strictly lower timeframe. Equal is not allowed; higher is not allowed.

• Window timeframe less than slot timeframe. A window shorter than the slot's own bar is not a meaningful accumulation range. The window must be at least as long as a slot bar.

These are not failures. They are the script refusing to compute something that would not be honest. If you hit one, the fix is to change the configuration so the constraint holds. Troubleshooting lists the specific messages and the single-step fixes.

What the defaults can and cannot claim

The defaults — 5m / 15m / 60m at equal weights, D window in Session mode, 1m/1m/5m Lower TF, Pressure Sensitivity 1.50, Wick Weight 0.20, CVD Length 3 SMA, Signal Length 3 SMA, Master Smoothing off, On Bar Close on — are a reasonable starting point for intraday reading on liquid symbols. That is what they claim.

They do not claim:

• To be correct for every instrument. A weekly chart, an illiquid altcoin, an overnight futures session, a session-traded equity after hours — each of these wants something different. Defaults are a starting point.

• To be the "best" configuration. There is no best. Any page that described a specific configuration as optimal would be doing something this pack explicitly refuses to do.

• To produce a particular read. The defaults plus your instrument's character produce a read. Change either and the read changes. The defaults cannot commit to a specific picture.

Run the defaults for a session on an instrument you know. Watch what the pane does. Decide what you want to change based on what you see, not based on an abstract belief about what a CVD tool ought to look like.

What the base trim does not do

Naming these explicitly rather than leaving a reader to assume.

• No cross-symbol slots. Every slot runs on the chart's ticker. Cross-asset studies (spot vs perp, index vs constituent) live in the CTX trim.

• No Power User MA parameters. No ALMA Offset, no KAMA or FRAMA Fast/Slow, no Jurik Phase or Power, no Laguerre Alpha, no VAMA Volatility Length. The base trim uses the Axiom MA library's family-level defaults. Readers who need to shape those MAs beyond family choice should reach for the CTX trim.

• No per-slot On Bar Close. The base trim exposes one global On Bar Close? switch that affects all three slots together. Per-slot repaint control is in the CTX trim.

• No crossing-at-level alerts. Overbought and Oversold are visual guides. The pack does not fire an alert when the line crosses them. Alerts is explicit.

• No divergence alerts. Divergence against price is a qualitative read the pane does not try to encode.

• No flip-only alerts. The alerts are state descriptors; any flip detection lives on the receiving side.

If an item on this list is a requirement for your process, the base trim is not the right choice. That is information, not an apology. Different trims for different jobs.

Liquidity and data-quality dependence

The pane's output is a function of the data underneath it. Three practical consequences.

• Thin volume at bar close is normal. Some instruments have brief volume dry spells inside an otherwise active day. The estimator will produce reads during those stretches, but the reads have less evidence behind them than reads taken during active hours. Expect the pane to be easier to trust during the parts of the day that the instrument itself is easier to trust.

• Holiday sessions, half-days, and illiquid hours. These produce unusual bar shapes and often trip the estimator's reading of participation. The pane is not malfunctioning during these stretches; the underlying data is unusual, and the read is telling you about unusual data.

• Data feed differences across brokers and exchanges. Two versions of the same symbol can carry slightly different OHLCV numbers on a given bar. The pane will reflect those differences. If you run the same configuration on two different feeds of the same instrument, small differences in the pane read are expected.

Pressure that persists against price

Worth repeating even though the Visuals page covers it. It is the single most common place where a trader decides the pane is wrong.

The pane reports participation-weighted pressure. Price reports the outcome of pressure against opposing forces — opposing participation, resting liquidity, market-making absorption. Those are not the same thing. A market can rally for an hour against selling participation before the selling overwhelms it. A market can sell off for a day against buying participation before the buying gives up.

When the pane and price diverge, the pane is not wrong. It is reporting something different from what price is reporting. The information is often useful: sustained pressure running counter to price is sometimes the leading edge of a reversal, and sometimes the leading edge of a long grind against the pressure that eventually succeeds. The pane does not tell you which.

Traders who expect price confirmation in real time will be frustrated. The frustration is not the tool's fault; it is a mismatch between what the tool reports and what the reader wanted it to report.

Honest trust budget

A practical frame for how much weight to place on the pane in a specific decision:

• High-weight contexts. Liquid instrument during active hours, default or carefully chosen configuration, multi-timeframe alignment present, pane read matches what you see in price structure. Place weight proportional to the rest of your confidence in the setup.

• Moderate-weight contexts. Liquid instrument, but either the configuration is still being calibrated, the timeframes are spread, or the pane read sits against price. The pane is information; it is not the center of the decision.

• Low-weight contexts. Illiquid instrument, post-reset calibration period, extended quiet-window pinning, persistent fallback to single-bar estimation, holiday or reduced-session hours. The pane is still telling you something honest; the something it is telling you is about a stretch of data that does not support strong reads.

The instrument does not know which bucket a given moment belongs in. You do. The pack will keep reminding you to use that knowledge.

A working checklist before a high-weight decision

When you are about to lean on the pane for a decision that will move real money, run through these five questions. None of them is optional, and a "no" on any of them is not a veto — it is context you should know you are spending trust on.

1. Is the instrument liquid enough right now that intrabar data is populated? Recent bars almost always are. Historical bars beyond the intrabar window are not. Sparse symbols may not be at any point. If you are reading a historical replay, the answer is usually that intrabars are missing and the line is in fallback mode.

2. Is the configuration one you can describe in a sentence? If you cannot reconstruct why a slot is in Rolling versus Session, or why a weight is at 20 instead of 33, the pane is answering a question you did not ask. Fix the configuration or do not trust the answer.

3. Is the read you are taking inside the window the pane can support? A pane read on the first three bars of a session is calibration, not information. A pane read during a known low-volume stretch is thin. A pane read on an instrument whose feed does not carry intrabar data is crude.

4. Does the blend match the stack? If the three slots are spread and the blend is near 50, the blend is reporting cancellation, not balance. Do not let the compressed headline drive the decision.

5. Have you checked against price? Pressure and direction are different reads. If they disagree, the pane is not wrong, but the weight of the read has to account for the disagreement.

A reader who runs through these five quickly will waste two minutes on a day the pane is fine and save their account on a day it is not. That is the trade the checklist is optimizing for.

Related pages

• MTF and Repainting for the timing model, the lookahead posture, and the fallback behavior.

• For the Geeks for the participation model at mental-model depth.

• Troubleshooting for specific symptoms and fixes, including what to do when a guardrail runtime error fires.

• Visuals and Logic for the ambiguous-middle reading that most often produces over-trust on this pane.