MACD Histogram: A Gap Between Two Lines
The MACD histogram plots the distance between the moving average convergence divergence line and its own nine-period signal line. Shrinking bars mean those two lines are converging, which is a statement about the indicator rather than about the direction of price.
How it works
The MACD indicator has two lines. The main line is a fast exponential average minus a slow one — conventionally 12 minus 26. The signal line is a nine-period exponential average of that.
The histogram is the first minus the second. Each bar is the vertical distance between the two lines at that point. Nothing else is in it.
Count the layers and the tool’s distance from price becomes clear. Price, then two moving averages, then their difference, then a smoothing of that difference, then the gap between those two. The histogram is four transformations away from a close, and each one costs a little more responsiveness and a little more meaning.
The two readings, and what each actually says
The histogram crosses zero exactly when the two lines cross. Those are not two signals. They are one event displayed two ways, and a chart showing both is showing the same information twice — worth knowing before treating agreement between them as confirmation.
The second reading is bar size: growing bars mean the lines are separating, shrinking bars mean they are converging. This is where the common misreading lives. Shrinking bars are frequently described as momentum fading, which sounds like a statement about price and is a statement about two averages getting closer.
Price can rise steadily while the histogram shrinks the entire time. A trend that advances at a constant rate has a stable gap between its fast and slow averages, and any steadying at all pulls them together. Nothing is weakening; the acceleration has stopped, which is not the same thing.
And the histogram has no bounds. Unlike the relative strength index or the stochastic, which are scaled 0 to 100, this is in the instrument’s own price units. A reading of 0.8 means nothing on its own — not across instruments, not across timeframes, and not across different price levels on the same instrument.
In practice
On this site’s shared 576-bar history, bar-to-bar changes in the histogram correlate at 0.71 with a 14-period relative strength index, 0.66 with the commodity channel index and 0.66 with 10-period momentum. It is not measuring something the rest of the oscillator family is missing.
Volume is absent from every layer. If you want an input that can disagree with the price side of the chart, it has to come from participation or from a different timeframe, not from another arrangement of closes.
Aggregating the chart changes the histogram completely, because 12 and 26 periods cover a different span of market. The settings are conventional, not derived, and they were chosen for a daily chart in an era of shorter trading weeks.
A gap produces the biggest bars you will see. The fast average moves much more than the slow one, the gap between them widens sharply, and the histogram registers maximum momentum for a move in which nothing traded.
It contains no price. Exiting “when the histogram turns” means exiting at whatever price happens to be trading when a derived quantity changes sign, which is not a stop and cannot be sized against.
Each zero cross traded costs 2% of a typical bar’s range in round-trip costs on this history, and a histogram oscillating around zero in a quiet market crosses repeatedly.
And the order book contains no histogram. Whatever the bars do, the market is not responding to them — other people watching the same conventional settings is the only mechanism available, and that is a weaker effect on a derived quantity than on a visible price level.
What the MACD histogram is not
It is not a momentum measurement of price. It measures the distance between two smoothings of price, which is a related but distinct quantity.
It is not a second signal alongside the line cross. Zero-crossings and line-crossings are the same event.
It is not comparable across charts. No fixed scale means no transferable levels.
And it is not independent of the other oscillators. 0.71, 0.66 and 0.66 against three of them on this data.
When it fails
In a range it oscillates around zero and crosses constantly. Every crossing looks like a signal, and the bars either side are small, so there is not even a size reading to filter them with.
The second failure is reading shrinking bars as weakness. A steady advance produces shrinking bars. Exiting on that basis means exiting trends that are behaving exactly as trends do.
A third is comparing histogram heights over time. Without a fixed scale, a bar twice as tall as last month’s says something about volatility as much as about direction.
A fourth is treating the standard 12/26/9 as meaningful. Those numbers are conventional. They were not derived from anything and they are not tuned to your instrument or timeframe.
And a fifth is the four-layer distance from price. Every transformation adds lag and removes information, and by the time you are reading the gap between a smoothing and a smoothing of that smoothing, the connection to what actually happened is thin.
The original data
Bar-to-bar change correlations on this site’s shared 576-bar history: the histogram against the
relative strength index 0.71, against the commodity channel index 0.66, against 10-period momentum 0.66,
against the stochastic 0.61. The full matrix — every oscillator pair, all between 0.53 and 0.87 — is in
research/series-measurements.json, produced by site/measure_series.py.
No pair in that matrix falls below 0.53, and that is the finding. Five oscillators built by different people at different times, from the same closes, and the least similar pair still moves together on more than half of all bars. The practical consequence is that a chart with three oscillators on it has roughly one oscillator’s worth of information and three times the screen clutter — and the way to find out which one to keep is to compute this matrix on your own instrument rather than to argue about which is best.
Related
MACD is the indicator the histogram is derived from. MACD crossover is the event the zero line marks. And MACD divergence is the comparison most often read off these bars.
The histogram is the part of this indicator I used longest without understanding, because shrinking bars feel like weakness. They are not - they are two smoothed lines getting closer, which happens on the way into a turn and also happens when a trend simply steadies.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.