Breadth · Construction
Reading a New-High / New-Low Chart
A new high is a price above the highest price of the last fifty-two weeks, a window with two ends. Knowing which end moved explains most of what looks surprising on these charts, including the long silences after a decline.
How the series is built
Each session, an exchange reports how many issues touched their highest price of the last fifty-two weeks and how many touched their lowest. Those two counts are the raw material, and they are usually presented in one of three forms.
The two counts side by side, which is the most informative version and the least common, because it is the only one that shows both counts being large at the same time, a state the other two forms average away.
The difference, new highs minus new lows, sometimes accumulated into a running line. Readable, and a count: it grows with the size of the exchange list, so it cannot be compared across decades.
The percentage, each count divided by issues traded, which is bounded and comparable. This is the form to use for any historical work, and the high-low index is one standard way of expressing it.
The window makes the pattern
A new high can be produced two ways: today’s price rises above the previous best, or the previous best drops out of the back of the twelve-month window. The second mechanism requires nothing to happen today at all, and it is responsible for the most conspicuous feature of these charts.
This has a practical consequence for any statement about new highs: it is incomplete without saying where the twelve-month window currently starts. In the year following a sharp decline the count is suppressed by construction, and a commentary noting that "new highs remain unimpressive" is describing the calendar.
The same mechanism operates in reverse and is less often noticed. When a period of sharp decline finally ages out of the window, new highs can expand rapidly across the list without any change in behaviour, the comparison simply got easier. Both movements are the back end of the window rather than the market, and both are entirely predictable in advance from the price history.
The asymmetry between the two counts
New highs and new lows are not mirror images, and treating them symmetrically, one threshold applied to both, mismeasures one of them.
| Property | New lows | New highs |
|---|---|---|
| How the count expands | Suddenly. Forced selling is synchronised, so a large share of the list can register within a few sessions. | Slowly. Buying is not forced, so the count accumulates over weeks. |
| Where the extremes sit | At market lows, and they are the most reliable extreme reading in this field. | Not at tops. Tops are marked by a narrowing count, not a spike. |
| What the window does to it | A prior low in the window suppresses the count in the same way. | Suppressed for up to a year after a peak, as the figure above shows. |
| Most common misreading | Treating an expansion as a forecast rather than as a description of selling under way. | Reading a low count during a recovery as weakness. |
The second row is the one worth carrying away, because it recurs across this whole reference: the extreme readings in every breadth and volume measure cluster at lows, not at highs. Tops are quiet, and they are identified (when they are identified at all) by participation narrowing rather than by anything spiking.
When both counts are large at once
The most interesting state these charts can show is one that the difference form hides completely: a session with many new highs and many new lows. Arithmetically it means the list has split, and in practice it usually means a rotation, money leaving one kind of issue for another, rather than a directional market move.
Several published indicators are built on this condition, the Hindenburg Omen being the best known. The common weakness of all of them is sample size rather than logic: the condition is rare, which is what makes it interesting and also means the evidence for what follows it rests on a handful of episodes. It is a reasonable thing to notice and a poor thing to act on with confidence.
What to check before quoting a figure from one of these charts
Which lookback. Fifty-two weeks is the convention; a chart using a quarter produces much larger counts that are not comparable with anything published.
Count or percentage. A raw count cannot be compared with an era when the exchange listed half as many issues, which is the same scaling problem the McClellan comparability page quantifies.
Which issues. Closed-end bond funds and preferred shares reach new highs and lows together on rate moves, and multiple share classes double-count one company. A common-stock-only series answers a different and usually better question.
Where the window starts. The single most useful check, and the one nobody makes. If a large price move sits inside the last fifty-two weeks, the count you are looking at is partly a statement about that move rather than about today.
Frequently asked questions
What is a new high, exactly?
A price above the highest price of some lookback window, almost always the last fifty-two weeks. That definition is doing more work than it appears to, because the window has two ends: a new high can be reached either by today’s price rising or by a high price dropping out of the back of the window. Half the surprises on these charts come from the second mechanism, and it is a property of the measure rather than of the market.
Why do new highs disappear for months after a decline?
Because the decline itself is still inside the twelve-month window. While the pre-decline peak remains in the lookback, an issue has to exceed that peak to register a new high, so a market can recover strongly and still produce almost no new highs, purely because the bar it must clear is the old top. New highs return in quantity only when either the price exceeds the old peak or the peak ages out of the window, whichever comes first.
Does that mean the absence of new highs after a crash is not a warning?
Correct, and it is one of the most commonly misread features of these charts. During the recovery from a sharp decline, a low new-high count is arithmetic, not weakness. The reading becomes informative again once the window has cleared the old peak, which is why any statement about new highs should be accompanied by where the twelve-month window currently starts.
Should I use the difference or the percentage?
The percentage, for anything spanning more than a few years. A raw difference between new highs and new lows is a count, and the number of issues on the exchange has changed by a factor of two or more across the decades people want to compare, so the raw form grows with the list. Dividing by issues traded gives a bounded, comparable series, which is the only version in which an old extreme means anything.
What does it mean when both counts are high at once?
That the market has split: some issues are reaching twelve-month highs while others reach twelve-month lows on the same day, which usually indicates a rotation between sectors or between rate-sensitive and cyclical issues rather than a directional move. Several published indicators are built on exactly this condition, the Hindenburg Omen among them, and their common weakness is that the condition is uncommon enough that the evidence for what follows rests on very few episodes.
Which issues distort the count?
The same ones that distort every breadth series: closed-end bond funds and preferred shares, which reach new highs and lows together on interest-rate moves and can produce a large count with no equity-market content. Multiple share classes double-count a single company. Where a common-stock-only version of the data is available it is the one to use, and where it is not, the count should be read as being about the listing rather than about companies.
Does a new low mean more than a new high?
The two are not symmetric, and it is worth knowing why. New lows expand suddenly and violently, because forced selling is synchronised and indiscriminate; the whole list can hit lows within a few sessions. New highs accumulate slowly, because buying is not forced. So a spike in new lows is a much sharper event than a spike in new highs, and any threshold applied symmetrically to the two sides is mismeasuring one of them.
How long a lookback should the chart use?
Fifty-two weeks is the convention and the one every published series uses, so it is what to use if you want your figures to be comparable with anyone else’s. Shorter windows — a quarter, or a month — produce a far more responsive series with a much larger count, and they lose the property that makes the twelve-month version meaningful: a fifty-two-week high is a statement about a full annual cycle of the instrument’s trading, and a twenty-day high is barely a statement at all.