Market breadth
Advance/Decline Ratio, the Formula and How to Read It
The simplest breadth measure there is: how many issues rose against how many fell. It answers a question a capitalisation-weighted index cannot: whether the whole market moved, or a handful of large names carried it.
The advance/decline ratio formula
The formula is one division, and how to read it is almost entirely a question of what the two counts contain, advancing issues against declining issues, taken from one exchange's list for one session.
A/D ratio = advancing issues ÷ declining issues, taken from one exchange for one session. A reading of 1.0 means as many issues rose as fell. Unchanged issues are excluded, which matters on thin sessions where they can be a large share of the list.
Two close relatives use the same two numbers and are frequently confused with it. The A/D line accumulates the daily difference, advances minus declines, into a running total, converting the same data into a trend. The breadth ratio some platforms plot is advances divided by advances plus declines, which bounds the result between 0 and 1 and is the form Zweig's thrust uses. They are not interchangeable, and a threshold quoted for one will be meaningless applied to another.
How to read it
How to read the ratio starts with what the formula discards: every advancing issue counts once whatever its size, and nothing in the number says how much was traded.
A single day's ratio is a weather report. Its value is comparative: against the index's own direction, and against the run of recent sessions. Three readings are worth knowing.
Divergence
The index makes a higher high while the ratio spends more and more sessions below 1.0. Fewer issues are participating in each advance, which means the index is being carried by its heaviest constituents. This is the most documented use of breadth data and the reason it survived into every serious market-internals dashboard. It is a condition, not a timing signal: narrow markets can grind higher for months.
Extremes at lows
Readings under 0.2 mean the selling stopped discriminating. Indiscriminate selling is characteristic of forced liquidation rather than of considered repositioning, and it clusters near lows for that reason. The subsequent day matters as much as the extreme itself: a washout followed by a ratio above 2 has a different character to one followed by another 0.4.
Thrust
A sustained move from broadly negative to broadly positive participation (the ten-day average crossing from below 0.40 to above 0.615 within two weeks, in the advances-over-advances-plus-declines form) is rare and has historically clustered near the start of durable advances. Rare is the operative word: it is not a signal you can build a routine around.
Where the measure came from
Breadth counting predates most of technical analysis. Exchange bulletins published advancing and declining issue counts long before anyone plotted an oscillator, for the ordinary reason that the numbers were easy to tally by hand from the day's sheets and answered a question a broker was asked constantly: was it a good day generally, or a good day for a few names?
The ratio became analytically prominent in the mid-twentieth century, when the concentration of American indices made the distinction matter more. Its most cited use is the divergence of the early 1970s, when breadth deteriorated for many months while the largest names carried the headline index into 1973, the pattern that made breadth analysis part of the standard toolkit. That episode is also the standing warning about it: the divergence was legible for so long beforehand that acting on its appearance alone would have meant sitting out a substantial advance.
What "counting" quietly assumes
Every issue counts once, and that equal weighting is the whole point, but it carries three assumptions worth making explicit, because each one is a way the reading can be technically correct and practically misleading.
That every issue is a company
On the NYSE it is not. The list carries closed-end funds, preferred shares, structured products and exchange-traded vehicles, and a large subset of them respond to interest rates rather than to earnings. On a day when the bond market moves sharply, several hundred line items can advance or decline together for a reason that has nothing to do with equity participation, and the ratio reports it as breadth. This is the single strongest argument for a common-stock-only count, and it is why thresholds carried forward from earlier decades describe a differently composed exchange.
That the issues traded
An issue that printed one hundred shares counts exactly as much as one that printed twenty million. In the aggregate this rarely matters, because the thinly traded tail is directionally random. It matters a great deal on quiet sessions, holidays and half days, when the actively traded subset shrinks and the tail becomes a larger share of the count.
That direction is meaningful for each of them
Unchanged issues are dropped from both sides of the division. On a busy session that is immaterial; on a quiet one, where a third of the list may be unchanged, the ratio is computed from a much smaller sample than the exchange's issue count suggests. Where a provider publishes the unchanged figure, an unusually high one is a reason to discount that day's reading rather than to explain it.
Intraday, and one level down
The ratio exists intraday and is a different animal there. Advances and declines are measured against the previous close, so in the first minutes of a session the count is almost entirely a description of the opening gap. Extreme readings are routine before any meaningful trading has occurred, and they settle as the session fills in. An intraday breadth figure taken early describes the open; taken at midday it starts to describe the day.
The same arithmetic applied to one sector answers the question the market-wide ratio asks of the index, one level down: was this sector's move general, or one or two large members? This is the reading that most often disagrees usefully with the headline. A market can print a healthy 1.8 while a single heavyweight sector is negative underneath it, and only the sector-level count shows that.
A worked reading
Suppose the index closes at a new high, the ratio prints 0.9, and the ten-day average of the ratio has been drifting down for three weeks. Three statements follow, and only the first two are supported.
- Supported: more issues fell than rose on a day the index rose, so the index gain came from its heavier constituents.
- Supported: this has been the pattern for three weeks, so it is a condition rather than a one-day artefact: which rules out the rebalance and futures-roll explanations that would apply to a single session.
- Not supported: that the advance is about to end. Narrow markets have ground higher for many months, and the same configuration has resolved upward often enough that treating it as a countdown is not defensible.
What the reading legitimately changes is exposure and expectation, not direction: an advance carried by fewer issues is more fragile to a shock in one of them, and it is a reason to check what the volume and the 52-week extremes are doing rather than a reason to act on its own.
Where it misleads
| Situation | What goes wrong |
|---|---|
| Exchange composition | Closed-end funds, preferreds and ETFs on the NYSE list move together on rate news, pushing the ratio to extremes for reasons unrelated to equities. |
| Historical thresholds | Levels drawn from an earlier decade describe a differently composed exchange; the same number does not mean the same thing. |
| Index rebalance day | Mechanical buying and selling across a whole list distorts the count with flows that carry no opinion. |
| Holiday or half session | Thin trading leaves many issues unchanged, and excluding them from the denominator exaggerates whatever ratio remains. |
| Single-day reading | Used alone it is noise. The divergence, not the day, is the information. |
Volume changes the reading
The ratio counts issues, never the size behind them, which is its strength and its blind spot in equal measure. A session where 2 000 issues rose and 1 000 fell looks identical whether the advances traded heavily or barely at all. Pairing it with up-volume against down-volume, or with TRIN, which combines both into one number, tells you whether broad participation was also committed participation. Broad and thin is a different market from broad and heavy.
Frequently asked questions
What counts as an advancing issue?
Any listed security that closed higher than its previous close, counted once regardless of size. That equal weighting is the entire point: a ratio of 2.0 means twice as many things went up as went down, and says nothing about whether they were the largest companies or the smallest. It is the deliberate complement to a capitalisation-weighted index, where a handful of names can carry the whole number.
What is a high or low reading?
On a broad exchange, days between roughly 0.7 and 1.5 are ordinary. Above 2 the session was broadly positive; above 4 or 5 is rare enough that it usually accompanies a policy surprise or a reversal off a washout. Below 0.5 the selling was indiscriminate, and readings under 0.2 — where almost nothing rose — cluster near capitulation lows rather than near tops. The asymmetry is real: fear is more synchronised than optimism, so extreme low readings appear more often than extreme high ones.
Why smooth it over ten days?
A single session is noise: one heavy futures roll or an index rebalance can distort the count. A ten-day average of the ratio, or of advances divided by advances plus declines, turns it into a measure of participation over a fortnight of trading, which is the horizon at which breadth divergences have historically been legible. Zweig's breadth thrust uses exactly this window, a ten-day average moving from below 0.40 to above 0.615 within ten sessions, which only works because the smoothing suppresses single-day artefacts.
Is the NYSE ratio still comparable to its own history?
Not straightforwardly, and this is the most important caveat on the page. The NYSE list now includes a large number of interest-rate-sensitive issues (closed-end funds, preferred shares, ETFs), that move together on rate news and are not operating companies. Their presence compresses the ratio toward the extremes on bond-driven days. Historical thresholds drawn from the 1960s or 1980s therefore describe a differently composed exchange, which is why many practitioners now prefer a common-stock-only count.
What happens to unchanged issues?
They are excluded from both sides, which is harmless on a busy session and distorting on a quiet one. If a third of the list closes unchanged (common on a half session, or in a market where many listed issues barely trade) the ratio is computed from a much smaller sample than the exchange’s issue count implies, and a handful of moves can swing it. Some providers publish the unchanged count alongside; where it is available, a session with an unusually high unchanged figure deserves less weight.
Should I use the ratio or the net difference?
They answer different questions and the choice is not cosmetic. The ratio divides, so it is scale-free: 1 500 advances against 750 declines gives 2.0 whether the exchange lists 3 000 issues or 6 000. The net difference subtracts, so it grows with the size of the list and is not comparable across exchanges or across decades. Use the ratio to compare, and the net, accumulated, when you want a trend, which is what the advance/decline line is.
Does the ratio work intraday?
It exists intraday and behaves differently enough to be treated as another indicator. Advances and declines are computed against the previous close, so early in a session the count is dominated by the opening gap and can be extreme before any real trading has happened. It settles as the session progresses. A reading taken in the first fifteen minutes describes the open, not the day.
How does it relate to TRIN?
TRIN, the Arms Index, is the advance/decline ratio divided by the up-volume/down-volume ratio. It exists precisely because counting issues ignores the size behind them: if breadth is positive but the volume is concentrated in the decliners, TRIN rises above 1 and says the participation was broad but not committed. Reading the two together separates a broad-and-heavy session from a broad-and-thin one, which the ratio alone cannot do.
Can I use it on a single sector?
Yes, and it is underused that way. A sector advance/decline ratio answers whether a sector’s move was general or driven by one or two large members, the same question the market-wide ratio asks of the index, one level down. It is particularly useful when the headline breadth looks healthy: a market can show a ratio of 1.8 while a single heavyweight sector is quietly negative underneath.
What is the highest reading ever recorded?
Depends entirely on the exchange, the era and whether unchanged issues were excluded, which is why quoting a record is less useful than it sounds. Extreme upside days above 10-to-1 occur, typically off a washout low or on a policy surprise, and downside days below 1-to-20 occur in genuine panics. The specific number matters less than its rarity: any reading beyond roughly 5-to-1 in either direction is a several-times-a-year event rather than a monthly one.
Why do practitioners smooth it rather than use the raw figure?
Because a single session contains a great deal that is not information: an index rebalance, a futures roll, a large basket trade. Smoothing over ten days averages those out and leaves a measure of participation over a fortnight, which is the horizon at which breadth divergences have historically been legible. The cost is that the smoothed series says nothing about today, which is the point: the raw figure is a weather report and the smoothed one is a climate reading.
Is a 1.0 reading neutral?
Arithmetically yes, behaviourally not quite. Because more issues rise than fall over long stretches in a market that trends upward, the long-run average sits slightly above 1.0 on most exchanges. A run of exactly-1.0 sessions therefore represents mildly weaker participation than the historical norm rather than a balanced market. The effect is small, and it is one more reason to read the figure against its own recent range instead of against a round number.
Does it help at market bottoms?
It is one of the few breadth measures that does. Extreme low readings cluster at lows rather than at highs, because indiscriminate selling (forced liquidation, margin calls, redemptions) hits the whole list at once, while narrowing at a top happens gradually. A reading under 0.2 does not say the low is in; it says the selling stopped discriminating, and the informative event is what the next two sessions do with that.
Which exchange should I take the data from?
Match it to what you are analysing, and prefer a common-stock-only count where the provider offers one. NYSE data compared against a large-cap index is comparing two different lists of companies; Nasdaq breadth is dominated by a much larger number of small issues and behaves accordingly. Mixing sources across a long history is the commonest way to produce a breadth chart that looks meaningful and is not.
Is it still useful now that indices are so concentrated?
More so, and this is the strongest argument for it. The gap the ratio measures (between what the largest constituents did and what the list did) widens as an index concentrates. A market where the ten largest names carry a substantial share of the capitalisation is precisely the market in which a weighted index and an equal-weighted count can tell opposite stories, and the ratio is the cheapest way to see it.