Volume is the count of shares traded and it has one property that trips everybody: it is not directional. Every trade has a buyer and a seller, so "more buyers than sellers" is arithmetically impossible. What high volume tells you is that a large number of holders changed their minds at once, which is information about conviction and about liquidity, and nothing at all about direction.
The useful volume readings are therefore comparative. A price move on volume far above its own recent average is a move that many participants took part in. A price move on thin volume may be one impatient order walking the book, which the microstructure work taught you to recognise. A breakout through a well-watched level on volume below average is a breakout most of the market did not join.
What are the main breadth indicators?
Breadth is the family of statistics computed across many securities rather than one price, and it is the part of this toolkit with the most defensible evidence base. Four constructions cover most of it.
Advertisement
The advance-decline line is the running cumulative sum of advancing issues minus declining issues, one number per session. It weights every listed name equally, which is exactly what a capitalisation-weighted index does not do, so the gap between the two measures how much of an index move is coming from its largest constituents.
The new-high-minus-new-low spread counts issues making fifty-two-week highs less those making lows. It is a stricter test than the advance-decline line, because a new high requires a full year of context rather than one session's direction.
The percentage of issues above their own two-hundred-day average is the cleanest single measure of participation, and it is bounded between zero and a hundred, which makes it comparable across time in a way the advance-decline line is not.
The McClellan oscillator takes the difference between a 19-session and a 39-session exponential average of net advances, which turns the cumulative line into a bounded momentum reading.
Why does breadth carry more weight than a chart pattern?
Two structural properties give breadth more weight than a pattern has. It aggregates hundreds of independent price series, so it carries far less idiosyncratic noise than any single chart. And it has almost no free parameters, so there is very little to curve-fit, with the advance-decline line having none at all.
A statistic with no knobs cannot have been tuned to the sample, and tuning to the sample is the commonest way a technical rule fools its author.
Worked Example 5 computes a breadth divergence in full. The honest caveat, which belongs in the same breath, is that breadth narrows regularly without a top following, and the divergence in the example runs for eight sessions before the index turns.
A breadth divergence is a change in the risk you are taking, not a forecast of a date.