Indicator library · Breadth
Advance/Decline Line
A running total of advancing issues minus declining ones. Its level means nothing; its slope, compared against the index it is drawn beside, is one of the oldest and most durable readings in market analysis.
The calculation
For each session, subtract declining issues from advancing ones and add the result to a running total: A/D line = previous value + (advances − declines). That is the whole formula. Unchanged issues are ignored.
The starting value is arbitrary and never matters. Because only the daily net is accumulated, beginning the series at zero or at ten million produces the same curve shifted vertically, which is why the vertical axis on an A/D line chart is usually left unlabelled and why comparing the level of two A/D lines is meaningless.
Reading it
There is essentially one reading, and it is comparative. Drawn beside the index it belongs to, the A/D line either confirms or contradicts what the index is doing. Confirmation is unremarkable and is the usual state. Contradiction (the index rising while the line falls) is the observation the indicator exists to produce.
The mechanism is worth stating plainly, because it explains both the value and the limits. A capitalisation-weighted index can rise on the strength of a handful of its largest members. The A/D line counts every issue once, so when the majority of the list turns down, the line turns down with it regardless of what the heavyweights do. The two series measure genuinely different things, and the gap between them is the signal.
The most cited historical example is the 1972 divergence ahead of the 1973–74 decline, when the American market's large names carried the index while breadth had already deteriorated. Its fame is also a warning: it is remembered because the resolution was dramatic, and the many divergences that resolved into nothing are not.
Where the numbers come from
Two counts per session, published by the exchange: how many issues closed above their previous close, and how many closed below it. Everything else on this page is arithmetic performed on those two numbers. That makes the indicator unusually cheap to compute and unusually sensitive to the definition of the list being counted, which is where most of the disagreement between data providers lives.
Three definitional choices change the series materially. The first is the universe: all listed issues, or common stock only. The second is the treatment of issues that did not trade, some vendors count them as unchanged, others omit them entirely. The third is whether the counts are taken at the close or at some point during the session, since an intraday reading of breadth is a legitimate thing to want and a completely different series from the one that gets accumulated into a daily line.
None of these are errors. They are decisions, and the only real mistake is to take a long history from one provider, splice a recent tail from another, and read the join as if it were a market event. If the shape of a line changes character on a date that has no market significance, suspect the data before the market.
A worked example
The first ten sessions of the series plotted above, showing every step. Advances and declines are in hundreds of issues; the net is their difference, and the line is the running total of the net. The line’s value in the final column is the only output.
| Session | Net (adv − dec) | A/D line | Index |
|---|---|---|---|
| 1 | +9 | +9 | 4180 |
| 2 | +6 | +15 | 4205 |
| 3 | -3 | +12 | 4198 |
| 4 | +8 | +20 | 4232 |
| 5 | +5 | +25 | 4251 |
| 6 | -2 | +23 | 4244 |
| 7 | +7 | +30 | 4278 |
| 8 | +4 | +34 | 4296 |
| 9 | -4 | +30 | 4288 |
| 10 | +3 | +33 | 4311 |
Notice what the table makes obvious and the chart does not: sessions three and six are negative nets inside an advance, and they lower the line without ending its rise. The slope is a property of the accumulation, not of any single session, which is why a one-day reversal in the line is not information and a three-week flattening is.
Variants that answer different questions
Several series carry the same name or a similar one and are not interchangeable. Knowing which one a chart shows matters more than any threshold quoted for it.
| Series | What it accumulates |
|---|---|
| A/D line | Advancing issues minus declining issues. One vote per company, regardless of size. |
| Common-stock-only A/D line | The same, with funds, preferreds and rate-sensitive vehicles excluded from the list. |
| Volume A/D line | Volume traded in advancing issues minus volume in declining ones, activity, not participation. |
| Percentage A/D line | Net advances divided by issues traded, then accumulated, corrects for a changing list size. |
| McClellan Summation Index | A running total of the McClellan oscillator, which is itself two averages of the same daily net. |
The common-stock-only variant deserves the extra effort where the data allows it. On a list containing several hundred closed-end funds and preferred issues, a single move in interest-rate expectations can push a majority of issues in one direction while the operating companies in the index do nothing. The line then records a breadth event that no equity investor experienced.
Using it alongside the index
In practice the line is read in one panel beneath the index it belongs to, on the same horizontal scale, and three questions are asked of the pair. Are they both rising? Has one stopped? Has the gap between them been widening for long enough that it is unlikely to be a week of rotation?
Two disciplines make the reading honest. Compare like with like, an NYSE line against a large-capitalisation index is comparing two different lists of companies, and the mismatch will produce apparent divergences that are nothing but a difference of membership. And date the divergence: writing down the session on which the line stopped advancing turns a vague impression into a fact that can later be checked, including against the many occasions when nothing followed.
Where it misleads
| Situation | What goes wrong |
|---|---|
| Reading the level | The origin is arbitrary. Only the slope and the comparison with the index carry information. |
| Non-operating issues | Closed-end funds and preferreds on the NYSE list move on rates, dragging the line for reasons unrelated to equity participation. |
| Divergence as a trigger | Divergences have run for many months before resolving, and some never resolve. It is a condition, not a signal. |
| Mismatched universe | An NYSE line compared against a large-cap index is comparing two different lists of companies. |
| Survivorship in long histories | Delistings enter and leave the count; very long A/D histories are not built from a constant universe. |
What volume adds
The A/D line counts participants and says nothing about how much they traded. Two sessions with an identical net of advances over declines can look nothing alike on the tape: one where the advancing names carried most of the day’s volume, and one where the volume went almost entirely into the minority that fell. The issue count cannot separate them, and the difference is usually the more interesting half of the story.
That is the gap the volume-weighted version of this measure exists to fill, and it is why breadth and volume are read together rather than treated as alternatives. Participation tells you how many names agreed; volume tells you how much conviction stood behind the agreement.
Frequently asked questions
Does the starting value matter?
No, and this is the property that confuses people first. The A/D line is a cumulative sum with an arbitrary origin, start it at zero, at a million, or at any other number and the shape is identical, because only the daily net is ever added. The level therefore carries no information at all. Only the slope, and the comparison against the index, mean anything.
What is an A/D line divergence?
The index makes a higher high while the A/D line makes a lower one. Because the line counts issues rather than weighting them by size, the divergence says the index advance is being carried by its largest constituents while most of the list is already falling. This is the reading the indicator exists for and the reason it survived from the 1920s into every modern market-internals screen. It is a condition, not a timing tool: divergences have persisted for many months before resolving, and some have not resolved at all.
How is it different from the advance/decline ratio?
They start from the same two numbers and answer different questions. The ratio divides advances by declines and describes one session; it is a daily reading that resets each day. The line subtracts declines from advances and keeps a running total, turning the same data into a trend that can be compared with the index over months. A threshold quoted for one is meaningless applied to the other.
Should I use the NYSE or a common-stock-only line?
Prefer common stock only where you can get it. The NYSE list contains a large number of closed-end funds, preferred shares and interest-rate-sensitive vehicles that are not operating companies; they move together on rate news and can push the line for reasons that have nothing to do with equity participation. Analysts have adjusted for this since at least the 1990s, and it is why a raw NYSE A/D line and a common-stock line can diverge from each other, never mind from the index.
How are unchanged issues handled?
They are discarded. Only advances and declines enter the subtraction, so a session in which most of the list closes exactly flat contributes almost nothing to the line even though a great many issues traded. This mattered far more before decimalisation, when a sixteenth of a dollar was the minimum increment and unchanged closes were common; on a modern tape, with prices quoted to the cent, genuinely unchanged closes are rare enough that the treatment is rarely the thing that distorts a reading.
What period should I plot it over?
Long enough that the slope is a trend rather than a wiggle, six months to two years on a daily series is where the indicator is usually read. Over a week the line is noise; over a decade the universe has changed underneath it through listings, delistings and mergers, so the far-left end of a very long history is not comparable with the right-hand end. Two years of daily data is the compromise most analysts settle on.
Can the line be computed weekly?
Yes, and a weekly line is quieter, but it is not the same series resampled. A weekly reading uses weekly advances and declines — issues up or down over the whole week — which is a different count from the sum of five daily nets, because an issue that rises Monday and falls Tuesday contributes to both daily counts and to neither weekly one. Never mix the two on one chart.
Is a rising A/D line during a falling index bullish?
It is the mirror-image condition and it is read the same cautious way: the majority of the list has stopped falling while the index, carried down by its heaviest members, has not. Historically this reverse divergence has appeared near the end of declines more often than at random, which is a much weaker claim than a signal. Treat it as a reason to look for confirmation elsewhere (in volume, in new lows drying up), rather than as an entry.
How does it relate to the McClellan oscillator?
They are built from the same daily net of advances minus declines and differ only in what is done with it. The A/D line accumulates the net without limit, so it is a trend. The McClellan oscillator takes the difference between a 19-period and a 39-period exponential average of the same net, so it oscillates around zero and says something about the rate of change of breadth rather than its direction. Reading both is reasonable; treating them as independent confirmation is not, because they share their only input.
Does it work on an individual stock?
No. The calculation requires a list of issues to count, so it exists only for a market, an exchange or a sector with enough members to count. Applying breadth logic to one instrument is a category error, though people do compute an A/D line for a sector index and compare it with the sector, which is legitimate as long as the member list is stable enough to be worth counting.
Where can I get advance/decline data?
The exchanges publish daily advancing and declining issue counts, and they appear in the market-summary tables of most financial data providers. Free historical series are patchier than they look: vendors differ on whether the count covers all issues or common stock only, and on how they treat issues that did not trade at all, so two providers can print different numbers for the same session. If a long history matters to you, take it from one source and keep it.
Should the line be indexed or normalised?
There is no need, because nothing about the level is being compared. Some analysts plot the line as a percentage of issues traded (net advances divided by total issues, then accumulated), which does correct for a growing or shrinking list and is a genuine improvement for multi-decade histories. Over the horizons where the indicator is actually read, the raw cumulative sum and the normalised version look almost identical.
Why is the vertical axis usually unlabelled?
Because a number on it would invite exactly the mistake the indicator cannot survive: reading the level. The origin is arbitrary, so the axis carries no meaning beyond the distances between points on it. Charting software that does label the axis is not wrong, but a reader who remembers "the A/D line is at 42,000" has remembered nothing at all.
What is the difference between this and the volume advance/decline line?
The volume version accumulates the volume traded in advancing issues minus the volume traded in declining ones, rather than counting the issues themselves. It answers a related but distinct question (where the day’s activity went, not how many names participated), and it can move opposite to the issue-based line when a few heavily traded names dominate the tape. On this site that measure is treated separately, under up/down volume.
Why has such a simple indicator lasted a century?
Because it measures something no capitalisation-weighted index can express, and it does so without a parameter to tune. There is no period to choose, no smoothing constant, no threshold, which means there is nothing to overfit and nothing that can be quietly optimised on past data to make a backtest look better than the future. Very few technical measures have that property, and it is the reason the A/D line survived the arrival of computers rather than being replaced by them.