Limitations & Trust Boundaries

Limitations and Trust Boundaries

This page maps the perimeter of what the oscillator can tell you, what it can only suggest, and what it cannot see at all. These are not disclaimers. They are the real edges of the tool's capability, and understanding them is what separates a trader who uses the oscillator carefully from one who over-trusts it in exactly the situations where trust costs the most.

The estimation boundary

The oscillator estimates volume delta from OHLCV candle data. It does not measure it. This is the most important limitation, and everything else in this page is downstream of it.

What OHLCV-based estimation can tell you: How each sub-bar's candle structure — body direction, close location within the range, wick imbalance — distributes across directional categories. When candle structure is clean and volume is meaningful, the estimate tracks the likely direction of volume commitment reasonably well. The participation model is more nuanced than a simple close-vs-open split, and the wick-aware classification adds information that the naive approach throws away.

What it cannot tell you: Who was buying and who was selling. Whether the volume was aggressive or passive. Where the limit orders sat. Whether a large player was accumulating or distributing. Whether the wick was genuine rejection or a liquidity sweep that got filled. The oscillator cannot distinguish between a market sell order and a limit buy getting filled at the same price — the candle looks the same from OHLCV data alone.

Where this matters most: When you look at a strong +80 reading and start reasoning about "buyers stepping in" or "institutional accumulation." Those are order-flow concepts. The oscillator does not have access to order-flow data. A +80 reading means the candle structure within the window has been heavily classified as bullish by the participation model. That is a real observation about candle structure. It is not a statement about market microstructure.

Here is how the slip happens in practice. You see the 1h slot at +75, price is grinding up, and you think "strong buying pressure — institutions are accumulating." The oscillator does support "the candle structure within this window has been predominantly bullish-classified." It does not support "institutions are accumulating." The price could be rising on light volume with small bullish candles that the model classifies as weakly bullish — enough to push the normalized reading high because the session has been one-directional, but backed by modest actual volume. Or the volume could be significant but driven by retail flow, not institutional activity. The oscillator cannot tell the difference. It reads candle structure. What you say about that reading to yourself — especially when you are already in a position and looking for confirmation — is where the estimation boundary gets crossed most dangerously.

Keeping this distinction clear under the pressure of a live session is the single most valuable thing you can do with this page.

For comparison: Footprint charts and depth-of-market tools work from tick-level trade data or order book snapshots. They can tell you the actual ratio of aggressive buys to aggressive sells on each price level. OHLCV estimation works from the four price points and total volume of each candle. The two data sources answer fundamentally different questions at fundamentally different resolution. This oscillator works well for what it works from — but pretending it works from something else leads to inferences the data cannot support.

The blending boundary

The blended CVD line is the most visible element in the pane and the one most likely to create false confidence.

What blending does: It produces a weighted average of the normalized CVD values across all enabled slots. When the slots agree, the blend faithfully reflects that agreement.

What blending hides: When the slots disagree. A blended reading of +15 could mean every slot is near +15, or it could mean some slots are at +60 and others are at -30. The blend averages the disagreement into a single middling number and presents it as if it were a unified reading.

Why this matters: The disagreement between timeframes is often the most informative state in a multi-timeframe analysis. A 5m slot flipping bearish while the 1h slot holds bullish is a specific, valuable observation about how pressure is distributing across timeframes. The blended line at +20 hides that observation behind a neutral-looking number. A trader who watches only the blend is getting less information, not more — despite the blend's appearance of being a higher-confidence composite.

The remedy is straightforward: Read the individual slots. Treat the blend as a convenience summary. When the blend shows a middling reading, check whether the underlying slots agree or conflict. The answer to that check often matters more than the blended value itself.

Post-reset instability

In Session mode, all slot values reset to zero at the window boundary. The normalization range also restarts, meaning the high and low used to scale the -100/+100 range are initially very narrow.

The practical problem: A modestly directional bar in the first few minutes of a new session can push the oscillator to +80 or -80 because the normalization range has not yet built enough data to put the reading in context. The number looks like a strong reading. The data behind it is thin.

How it manifests: If you are watching the oscillator at the session open, you may see sharp moves to extreme values that settle down within 15-30 minutes as the normalization range widens. If you are reviewing historical charts, post-reset bars may show apparently extreme readings followed by a "pullback" that was really just the normalization range expanding.

What to do about it: Give session-mode readings 15-30 minutes of data before treating them as reliable context. If a genuinely strong move starts at the session open, the oscillator will show it — but the early extreme readings should be weighted less in your judgment than the same readings mid-session. There is no mathematical threshold where "enough bars" makes the reading fully reliable. It is a matter of degree, and the direction is clear: more data, more context, more stable normalization.

Cross-ticker normalization

When two slots target different symbols, both produce -100/+100 readings — but the normalization is independent per slot. Each slot normalizes against its own window's range of cumulative delta.

What "range-comparable" means: A reading of +60 on two different slots means both are near the top of their own window ranges. Both are extended relative to their own history.

What it does not mean: That both are experiencing the same volume magnitude, the same absolute delta, or the same market force. A +60 on BTCUSDT — a market with billions of dollars of daily volume — and a +60 on a small-cap altcoin with a fraction of that volume represent fundamentally different things in absolute terms. The normalization makes them look equivalent on the screen because it scales each to its own range.

Where this creates risk: In the blended line. When the blend averages a +60 BTC slot and a +40 altcoin slot into a blended +50, it is mixing two normalized scores from markets with very different volume characteristics. The blended number looks clean, but it does not represent a volume-weighted composite of capital flow between markets. It represents a composite of relative extension within independent normalization windows.

What to do about it: If you use cross-ticker slots, treat the individual slot readings as independent observations about each market's internal pressure dynamics. The blend of cross-ticker slots is a convenience summary of how those relative extensions are averaging. Do not treat it as evidence of coordinated buying or selling between markets. Regime agreement across tickers is suggestive — "both are extended bullish" — but it is not confirmatory of a causal or capital-flow relationship.

OB/OS levels as context, not prediction

The overbought (+70 by default) and oversold (-70) reference lines are thresholds on the normalization range. When the blended CVD crosses above +70, it means the composite reading is in the top 30% of its current normalization range.

What this does not predict: Reversals. A reading above +70 in a trending session can persist for hours. The oscillator is designed to stay extended when pressure is consistently directional. Treating the OB cross as a fade signal without supporting context from price action, order flow tools, or structural analysis is one of the most common misuse patterns.

The misread in action: It is 10:45am, the blended CVD has been above +70 since 10:15, and the trader thinks "overbought for 30 minutes — this is due for a pullback." They short. The oscillator stays above +70 for the next hour because the session's volume is genuinely directional. The trader was not reading an overextension — they were reading a trend. The +70 level told them the composite was extended relative to the normalization range. It did not tell them the extension was about to stop. Extension during a trend is normal. It is the oscillator doing its job. Fading the number without corroboration from price structure, from a slot that has diverged, or from some other source of caution is the kind of decision that this page exists to prevent.

How to use it well: If you are already looking for reasons to be cautious about chasing a move, an extended OB reading is a relevant data point — not the reason to act, but a piece of context that says "the composite is stretched." If the blended line is above +70 and the shortest-TF slot is starting to decline while the longer-TF slots hold, you have something more specific: extension plus early divergence. That combination is worth more than the OB number alone. If the blended line is above +70 and all slots are still rising, the reading is telling you that estimated pressure has been consistently directional — that is information about persistence, not about imminent reversal.

What the zero line means — and does not mean

A reading near zero means estimated buying and selling pressure within the slot's active window are roughly balanced. The net delta is close to zero.

What it does not mean: That volume is absent. A session with heavy volume and roughly equal buying and selling pressure will produce a reading near zero, just like a session with almost no volume. If volume participation matters to your analysis, the oscillator cannot tell you about it. Check volume bars separately.

What it also does not mean in Rolling mode: That nothing is happening. In Rolling mode, the oscillator can drift toward zero because old directional bars are leaving the window, not because new bars are neutral. A declining reading that approaches zero may represent fading historical pressure, not current balance. The current bars could be adding directional volume that has not yet outweighed the rolling-out of older bars.

The silent fallback

When a slot requests intrabar data via Lower TF Precision and the platform cannot provide it (sparse history, symbol limitations, very low timeframe on certain exchanges), the indicator falls back to estimating delta from the slot bar's own OHLC and volume instead of from the sub-bar array.

This fallback is silent. The oscillator continues to produce values. The values are based on coarser input, which means the participation model has less to work with and the estimate is rougher. There is no visual indicator, no warning, and no way to tell from the chart alone whether a given bar was estimated from sub-bar data or from a single-bar fallback.

Why this matters for trust: You may be comparing the oscillator's performance on two instruments and find that it "reads" one better than the other. Part of that difference may be the silent fallback. On an instrument with deep, continuous data, the participation model gets many sub-bars to classify per slot bar — each contributing a directional assessment. On an instrument with sparse history or exchange-specific gaps, some bars may be classified from a single candle, producing a rougher estimate. The oscillator does not look different in either case. The numbers look just as confident. But the inputs behind them are not equivalent.

What to do about it: If you are working with instruments that have limited historical data, very low timeframes, or exchange-specific data gaps, be aware that the delta estimates on some bars may be rougher than others. If the oscillator's behavior on a specific instrument seems less responsive than expected — or produces readings that do not match what the candle structure seems to show — the silent fallback may be a factor. You can test this indirectly by changing Lower TF Precision: if adjusting it produces no visible change in the CVD trajectory, the sub-bar data may not be available and the fallback may already be active.

What the oscillator can reveal, what it can only suggest, and what it cannot see