Indicator library · Breadth
New Highs and New Lows
A daily count of how many issues reached the highest or lowest price of their past year. Slower than advance/decline data, harder to distort, and most informative in the case people overlook, when both counts are large at the same time.
The calculation
Two raw numbers per session, per exchange: how many issues traded at a 52-week high, and how many at a 52-week low. Three derived series are built from them.
- Net new highs: new highs − new lows, the form plotted below. Simple, and dependent on how many issues the exchange lists.
- High-low index: new highs ÷ (new highs + new lows) × 100, usually with a ten-day average. Bounded 0 to 100 and independent of list size, which makes it the version worth comparing across exchanges or across decades.
- Cumulative net: a running total of the net, read the same way as an advance/decline line: level meaningless, slope informative.
The rolling window deserves attention. A "new high" is measured against the previous 52 weeks, so the threshold moves every day as old data drops out of the back of the window. A year after a severe decline the comparison period contains that decline, and new highs become mechanically easier to reach, a step-change in the counts that reflects the calendar rather than the market.
Reading it
Expansion and contraction
A market making new index highs with an expanding number of individual new highs is broad. The same index highs with a contracting count is narrow, and narrowing has historically preceded weakness often enough to be worth watching, though, as with every breadth divergence, it can persist far longer than it is comfortable to sit with.
New lows expanding
The asymmetry of breadth data shows up here clearly: expanding new lows tend to lead index weakness more reliably than contracting new highs do. Selling is more synchronised than buying, so the low count moves first and moves harder.
Both at once
The case that gets overlooked. Large counts on both sides mean the list has split into two groups going in opposite directions, an internally divided market rather than a directional one. The Hindenburg Omen formalises this into a signal; treat the underlying observation as sound and the signal itself with the scepticism its historical hit rate deserves.
The high-low index, worked through
The bounded form is worth computing by hand once, because the arithmetic explains both its advantage and what it throws away. Four sessions, invented to isolate the effect:
| New highs | New lows | Net | High-low index |
|---|---|---|---|
| 120 | 20 | +100 | 85.7 |
| 12 | 2 | +10 | 85.7 |
| 200 | 180 | +20 | 52.6 |
| 10 | 9 | +1 | 52.6 |
The first two rows are identical to the index and wildly different on the net: a market with 140 issues at their yearly extremes is not the same market as one with 14. The third and fourth rows are the case that matters most, 380 issues at extremes in opposite directions is the internally divided market described below, and the bounded index reduces it to a reading barely distinguishable from a quiet, balanced session.
The practical rule follows from the table rather than from anyone’s preference. Use the index when comparing across exchanges, across decades, or against a period when the listed universe was a different size. Keep the raw counts in view whenever the question is whether anything is happening at the extremes at all, and always look at the two counts separately before reading any ratio built from them.
Where it misleads
| Situation | What goes wrong |
|---|---|
| Rolling-window artefact | A year after a crash, the comparison period contains the crash: new highs become easier to set for calendar reasons alone. |
| Raw counts across exchanges | The numbers depend on how many issues are listed. Use the bounded high-low index to compare. |
| Non-operating issues | Preferreds and funds set 52-week extremes on rate moves, not on business performance. |
| Newly listed issues | A stock with less than a year of history cannot set a 52-week extreme, so heavy IPO periods quietly shrink both counts. |
| Hindenburg Omen as a trigger | The signal fires far more often than the outcomes it is credited with predicting; the division it detects is real, the forecast is not. |
Frequently asked questions
What exactly counts as a new high?
An issue whose price today reached the highest level it has traded in the previous 52 weeks. The window is rolling, which produces an effect worth knowing about: as an old extreme drops out of the back of the window, the bar for a "new high" can fall without anything happening today. Counts taken a year after a crash are therefore easier to achieve than they look, because the comparison period includes the crash itself.
What is the high-low index?
New highs divided by the sum of new highs and new lows, expressed as a percentage and usually smoothed with a ten-day moving average. It bounds the reading between 0 and 100 and removes the dependence on how many issues the exchange happens to list, which makes it comparable across exchanges and across decades in a way the raw counts are not. Above 50 means new highs are dominating; the smoothing is what makes the series readable at all.
Why does it matter when both counts are high at once?
Because it should not happen. A healthy market produces many new highs and few new lows, or the reverse. A day where both are substantial means the list is splitting — one group of issues making new highs while another makes new lows — which is a market without a single direction. This condition is the core of the Hindenburg Omen, a much-discussed signal whose historical record is far weaker than its fame, but the underlying observation about internal division is sound even where the signal built on it is not.
Is it a leading or a lagging measure?
Slower than the advance/decline data and correspondingly steadier. A stock has to move a long way to set a 52-week extreme, so the counts do not react to a single session; they describe how many issues are at the edges of their own year-long ranges. In practice new lows expand ahead of index weakness more reliably than new highs contract ahead of it, the asymmetry that runs through most breadth data.
Are intraday extremes counted, or only closing prices?
It depends on the publisher, and the two versions are not interchangeable. Counting any trade at a new extreme produces consistently larger numbers than counting only closes at a new extreme, because a stock can touch a high early and close well below it. Neither is wrong. What is wrong is comparing a long history from one convention against recent data from the other, which produces a step in the series that looks like a change in market character.
Why can a stock not set a new high shortly after listing?
Because there is no 52-week window to compare against. Most publishers require a full year of trading history before an issue is eligible, so a period of heavy new listings quietly shrinks both counts without any change in behaviour among the companies already listed. In a year with an unusually active IPO calendar this is a real distortion, and it works in the direction of understating breadth.
Should I use net new highs or the high-low index?
The index for anything comparative, the net for reading a single market over a few months. Net new highs carry the size of the exchange’s list with them, so a count of 120 means something different on a list of 3,000 issues than on a list of 1,500, and the American exchanges have changed size substantially over the decades. The high-low index divides that dependence out and bounds the result between 0 and 100, at the cost of hiding whether the day was busy or quiet at the extremes.
What levels of the high-low index matter?
Fifty is the only level with a definition rather than a convention behind it: above it, new highs outnumber new lows. Readings that stay above about 70 or below about 30 for a stretch describe a market in which one side of the extremes has been dominant for weeks, which is a description rather than a signal. The smoothed version spends most of its life between those markers, so the extremes are rare enough to notice without being rare enough to act on.
Does the measure work on a single sector?
Yes, provided the sector has enough members for a count to be meaningful, a few dozen at minimum, and preferably many more. Sector-level new-high counts are useful precisely because they can contradict the market: a sector setting internal new highs while the market’s count contracts is the sort of rotation that index-level breadth averages away. The eligibility caveats still apply, and small sector lists are noisy.
How does this relate to the advance/decline line?
They measure participation at different distances. The A/D line counts every issue every day against yesterday, so it registers small moves and reacts immediately. The high-low counts register only issues at the boundary of their own year, so they ignore almost everything and react slowly. The two can disagree for weeks without either being wrong, and when they agree (both narrowing while the index rises) the observation is considerably stronger than either alone.