OBV in crypto is a running total that adds a candle's reported volume when its close rises, subtracts it when its close falls, and leaves the total unchanged when the close is equal. Use it to describe a specific market's volume-weighted closing direction. It does not measure net money entering crypto, identify informed buyers or establish that a divergence will produce a reversal. The venue, volume unit and completed-candle state belong beside every OBV claim.
A tiny change in the closing price can assign an entire candle's volume to the other side of the calculation. That makes OBV easy to reproduce and easy to overinterpret. Before reading the line as conviction, inspect what changed its direction.
What goes into the running total
For each bar, compare its close with the preceding bar's close. Add current volume for a higher close, subtract it for a lower close, and add zero for an equal close. This is the conventional calculation documented in TradingView's On Balance Volume guide.
The comparison is close to close. A candle that closes above its own open can still subtract volume if it closes below the preceding close. The size of the price change does not scale the contribution: a rise of 0.01 and a rise of ten both add all the bar's volume under this rule.
The running total has the units of the selected volume field. Binance's spot candlestick response separates volume, quote-asset volume and taker-buy fields; its spot WebSocket kline schema explicitly identifies v as base-asset volume. Binance spot candlestick fields and Binance spot kline stream schema establish those particular field meanings.
Do not silently substitute quote turnover for base quantity or splice a derivative series into a spot series. The volume confirmation guide covers unit selection and volume baselines in more detail. OBV's distinct question is how those reported quantities accumulate after the close assigns a sign.
Reconstruct six fictional candles
The following example uses invented completed hourly spot candles from one venue, base-asset volume and an initial OBV of zero. The close immediately before bar A is 100 quote units. This is a calculation fixture, not market evidence or a trading recommendation.
| Bar | Close | Base volume | Close versus previous | OBV change | Running OBV |
|---|---|---|---|---|---|
| A | 101.00 | 40 | Higher | +40 | 40 |
| B | 100.99 | 10 | Lower | −10 | 30 |
| C | 100.99 | 500 | Equal | 0 | 30 |
| D | 101.02 | 25 | Higher | +25 | 55 |
| E | 100.98 | 70 | Lower | −70 | −15 |
| F | 101.03 | 20 | Higher | +20 | 5 |
The contributions sum to 40 − 10 + 0 + 25 − 70 + 20 = 5. Total reported volume across the six bars is 665 base units. OBV finishes at five because the 500-unit unchanged-close bar contributes zero, while the remaining signed contributions mostly cancel.
Neither five nor 665 tells you net buyer demand. Every matched trade has both sides, and OBV assigns the whole candle's volume according to one closing-price comparison. A large unchanged-close bar may contain substantial movement and trading inside it; the endpoint rule discards that detail.
Now hold everything through E fixed and change only F's fictional close from 101.03 to 100.97. Its volume remains 20. The final contribution becomes −20, so OBV finishes at −35 rather than five. The 0.06 difference between the alternative closes changes the running total by 40 base units. This is the rule's arithmetic, not a measurement error.
Read divergence as a defined comparison
In the original table, close D is 101.02 with OBV at 55. Close F is higher at 101.03 while OBV is lower at five. Those observations create a higher-price, lower-OBV comparison. It raises a question about the volume behind that particular advance; it does not supply an entry price, reversal deadline or loss limit.
The same distinction applies to oscillator disagreement. The RSI divergence failure worksheet separates a confirmed comparison from a price trigger, invalidation and expiry, so a later reversal cannot retroactively qualify an earlier canceled setup.
For a reproducible divergence rule, state whether you compare closing-price peaks, candle highs or another series. Name the two observations and how they were selected. If a peak requires two subsequent bars to confirm it, the peak was not confirmed at its own timestamp. The look-ahead bias audit explains why using that later confirmation too early invalidates a simulation.
Selecting D and F after seeing a subsequent decline would be an observation chosen with hindsight. A research rule needs selection criteria before outcomes are inspected, plus a decision time and an expiry. An OBV divergence can persist while price continues in the same direction; the indicator's definition contains no mechanism forcing price to reverse.
Keep the starting point separate from the signal
Repeat the table with an initial OBV of 10,000 instead of zero. The sequence becomes 10,040, 10,030, 10,030, 10,055, 9,985 and 10,005. Every contribution is unchanged. The comparison between D and F is unchanged too.
This demonstrates why “OBV is positive” needs a seed definition. The zero crossing can move when the initial constant changes. A higher number on another platform need not mean stronger buying pressure; it may simply reflect a different history start or initialization.
An OBV change over the same interval avoids a constant seed offset, provided all underlying bars and fields match. It does not solve missing candles, revisions or different venue activity. Preserve the seed timestamp, volume units and history used to reconstruct the line. The candle-data quality worksheet offers checks for gaps, conflicts and duplicates that can alter cumulative calculations.
A live candle can reverse its entire contribution
Continue after F, where the last completed close is 101.03 and OBV is five. Suppose the next unfinished candle currently shows a close of 101.04 and volume of 600. Its provisional contribution is +600, giving a provisional OBV of 605.
If that same candle ultimately closes at 101.02 with final volume of 650, its completed contribution is −650 and completed OBV is −645. These are alternative updates from the same completed state of five. Adding both would count the candle twice.
Binance's spot kline stream exposes an x field indicating whether the kline is closed. That is a useful completion flag for that source, separate from whether the connection is alive. Binance spot WebSocket kline completion field.
A candle-close confirmation rule must wait for a finalized observation. An intrabar rule needs historical intrabar observations to test the same behavior honestly. A final candle archive cannot reveal every provisional OBV state. Check the market-data freshness record before interpreting a live reading.
Use a compact OBV decision record
| Field | Record before interpreting the line |
|---|---|
| Market identity | Venue, exact pair, spot or contract, and source |
| Input definition | Candle interval and boundaries, volume field and units |
| Calculation state | Seed, start timestamp, completed or provisional bar |
| Comparison | Named price observations, matching OBV values and selection rule |
| Failure condition | What observation would contradict the interpretation, and by when |
| Evaluation | Unseen data, costs, execution assumptions and tested alternatives |
One exchange's pair is one measurement scope. A composite needs an explicit construction method; adding raw lines with different units or seeds has no clear meaning. A plausible divergence can also fail when reported activity is unrepresentative, data are missing or the proposed trade cannot execute near the assumed price.
This article was prepared with AI assistance using the linked primary documentation. Its prices, volumes, seed test and live-candle alternatives are original hypothetical teaching examples. They explain a calculation and its limits; they do not establish a profitable OBV strategy or a particular KLINEO indicator implementation. Keep the observation distinct from the decision and its execution costs.