Joseph Granville: What OBV Counts and What It Misses

A stock closes one cent higher on five million shares. Its On-Balance Volume line jumps by five million, just as it would if the stock had closed five dollars higher on the same volume. Joseph Granville’s On-Balance Volume indicator makes a deliberately coarse distinction, and that distinction shapes every interpretation of it.

StockCharts’ OBV reference attributes the indicator to Granville’s 1963 New Key to Stock Market Profits. It explains the cumulative calculation, the arbitrary starting value and the possibility that unusual volume spikes distort the line.

The examples below are hypothetical and independently constructed. They examine the arithmetic directly, without treating a rising or falling line as proof of institutional activity or a forecast of a particular return.

Joseph Granville and the three possible daily entries

OBV adds the current session’s volume when its close is above the previous close. It subtracts that volume when the close is below the previous close. If the two closes are equal, OBV is unchanged. The size of the price change doesn’t weight the contribution.

Start a hypothetical series at zero with a reference close of 20.00. The next close is 20.10 on 100,000 shares, taking OBV to 100,000. A subsequent close at 20.05 on 150,000 shares takes it to minus 50,000. Another close at 20.05, this time on 200,000 shares, leaves the total unchanged.

I find the unchanged 20.05 close especially useful: 200,000 shares traded, yet the indicator records no movement. OBV assigns each session’s volume according to the closing-price comparison, so that 200,000-share session leaves the line unchanged.

The calculation can be checked without making any claim about motives. It doesn’t identify who initiated individual transactions, why they traded or how many institutions were involved. Those questions require information that a daily close and total volume don’t contain.

The same trend can have different OBV totals

Suppose one chart starts OBV at zero and another starts at one million. Applying the same subsequent price and volume observations leaves a constant difference of one million between their lines. Their changes over any shared interval are identical.

The sign of an absolute total can reflect the starting convention. Interpreting it as a judgement of the instrument’s strength would confuse that convention with subsequent behaviour. The relevant comparison concerns changes and turning points within a consistently calculated series.

Nor can raw totals be ranked sensibly across unrelated stocks. A company trading ten million shares daily can produce a much larger OBV movement than one trading 100,000, simply because the scale of its volume differs. The scale needs to be considered before comparing signal strength.

The Volume Rate of Change guide describes another way of expressing changes in activity. It answers a different question. Switching indicators changes the measurement, so the interpretation needs to change with it.

A rising price with a falling volume line

Consider three hypothetical closes after a 30.00 reference: 31.00 on 100,000 shares, 30.80 on 300,000 shares, then 31.20 on 100,000 shares. Price finishes 1.20 higher. OBV, starting at zero, finishes at minus 100,000.

The down-day subtraction of 300,000 outweighs the two up-day additions totalling 200,000. There is no negative volume on either up day. The falling cumulative line follows directly from the indicator’s sign convention.

I would describe this example as a disagreement between the price change and the signed volume total. Calling it confirmed distribution by institutions adds an identity claim that the three observations can’t support.

A trader studying volume around breakout candles might investigate whether this kind of disagreement affects later results. That needs a sample and defined exits. The divergence alone supplies neither a reversal date nor a probability that the next breakout will fail.

A large session can dominate weeks of observations

Imagine an up day with five million shares, followed by ten down days with 200,000 shares each. OBV rises by five million and then gives back two million, leaving a net gain of three million. Depending on the prices, the stock itself could have fallen materially during those ten sessions.

The large day’s influence remains in the cumulative series. It can make the line look resilient even while more recent observations have a different character. A reader who sees only the latest total may miss the dependence on that earlier spike.

A review could retain the full series while also recording the net OBV change over a fixed recent interval. That provides another view without deleting inconvenient data. Removing the five-million-share day after seeing the result would create a new indicator and a new test.

A spike also warrants a data check. Corporate actions, unusual events and the data vendor’s volume conventions can matter. The correct response depends on what happened. The observed bar and the explanation for it belong in separate fields of the review.

Intraday location and close-to-close direction differ

A hypothetical stock closes yesterday at 40.00. Today it opens at 42.00, reaches 43.00, falls to 40.10 and closes at 40.20. OBV adds today’s volume because 40.20 exceeds 40.00, even though the session closed well below its opening price and near its low.

That is correct OBV arithmetic. It also shows why the line can’t replace examination of the price bar. A positive contribution doesn’t say that the stock spent the session rising or finished near its best price.

The support and resistance guide gives a separate way to describe where the session finished relative to earlier levels. If 41.00 was a prior ceiling, the 40.20 close is back below it. The positive OBV contribution and the failed closing breakout can coexist.

Different indicators weight these details differently. Combining several of them doesn’t automatically create independent confirmation, since they may reuse the same price and volume inputs. Their formulas and the decisions they inform deserve attention before their agreement is counted as extra evidence.

Define divergence before inspecting the outcome

A hypothetical study might compare two price highs separated by at least five sessions and ask whether OBV’s value at the second high is lower. That is one possible definition. Another might compare the highest OBV value within a small window around each price peak.

The definitions can classify the same chart differently. If price peaks on Tuesday and OBV peaks on Thursday, an exact-date comparison and a surrounding-window comparison won’t necessarily agree. Neither is made correct by whichever produces the most attractive later chart.

The timing of a confirmed swing matters too. A peak identified only after several lower bars isn’t known on the peak date itself. A simulated decision must use the date when the identifying information becomes available.

For another Masters profile, see Ed Seykota. The OBV test here is an independent illustration. Precise definitions let a losing example remain part of the record instead of disappearing through reinterpretation.

Volume data establish the limits of the exercise

Suppose a platform supplies only activity from one venue while another provides a broader consolidated series. The daily totals can differ even when the closing prices match. OBV calculated from those totals can consequently show different changes and divergences.

A research note can state the instrument, data source, trading session and adjustment convention. For a session spanning midnight, even the closing boundary can determine whether volume is added or subtracted. Comparisons are easier to interpret when those choices remain consistent.

The same caution applies when replacing traded volume with a proxy. A count of quote changes isn’t identical to the number of shares or contracts transacted. A useful proxy may deserve study, but its interpretation must follow the data actually supplied.

These details are less dramatic than a chart showing divergence before a large reversal. They decide whether the calculation can be reproduced and whether apparently similar examples belong in the same sample.

A missing session deserves an explicit flag. If a data feed omits a down close with 400,000 shares, every later cumulative total differs by that contribution until the omission is repaired. The chart can still look smooth. A reviewer could check several daily additions and subtractions against the underlying records before interpreting a surprising divergence. This basic reconciliation protects the analysis from a problem that no additional indicator can diagnose reliably from the finished OBV line alone.

Books for studying Joseph Granville

New Key to Stock Market Profits is linked here as a paperback reprint, distinct from the 1963 original publication discussed above. Granville’s New Strategy is a separate hardcover title. Both are historical reading; neither book title establishes the reliability of the five-million-share example or any modern trading rule.

As an Amazon Associate, I earn from qualifying purchases.

Granville’s New Strategy of Daily Stock Market Timing for Maximum Profit. Hardcover.

(US) Granville’s New Strategy of Daily Stock Market Timing for Maximum Profit

New Key to Stock Market Profits. Paperback reprint.

(US) New Key to Stock Market Profits

Keep the arithmetic beside the interpretation

In the 31.20 closing example, the observable fact is that 300,000 shares were subtracted and 200,000 were added. That statement survives disagreement about what happens next. Granville’s indicator becomes easier to use critically when the calculation remains visible beneath the story attached to it.

Learn the pattern. Ride the trend. Keep the gains.

Educational content only. Not investment advice. Trading involves risk. You are responsible for your decisions.