Reference · Daily data
What a Market Summary Should Contain
A summary of a session needs five numbers, and the one in every headline is not among them. The index level says what the largest companies did; the other five say what the market did.
Why a specification rather than a summary
A daily summary is a description of one session, and a description of one session has a short life. Read a week later it is history; read a year later it is misleading, because the reader has no way of telling how much of it was true only that morning.
What does not expire is the specification: which numbers belong in a summary, what each one supports, and how to read them against each other. That is also the argument the whole site is built on, so it is worth stating in one place.
The five, in order
1 · How many took part. Advancing issues against declining ones, with the unchanged count published alongside. This is a census: every issue gets one vote regardless of size, which is exactly what an index cannot do. The anatomy page sets out which issues distort it, funds and preferred shares moving together on rate news, and multiple share classes counted twice.
2 · What the move cost in stock. Total volume as a ratio to its own twenty-session average, not as a share count. A ratio is comparable with last month and with 1998; a raw total is not, because listings and float have changed. Around 1 is ordinary, above 2 is genuinely busy, and above 5 means something specific and usually identifiable happened.
3 · Whether the extremes are crowded. New twelve-month highs and new twelve-month lows, reported separately. Both being large at once is the informative state and a difference hides it. State where the twelve-month window starts, too: after a sharp decline the high count is suppressed by arithmetic for a year, which the new-high page computes.
4 · How far the participants travelled. The average percentage range per issue, which is a volatility measure for the market rather than for one instrument, elevated through declines, compressed in quiet advances. Average the percentages rather than the point ranges, or the most expensive constituents dominate.
5 · Whether the money agreed with the count. Up volume against down volume. It matters only when it disagrees with the issue count, and those sessions are uncommon, the up/down volume page computes how uncommon on a synthetic series. When they do disagree, look for the one enormous issue before reaching for an interpretation.
Reading them together
| Configuration | What the summary should say |
|---|---|
| Index up, broad advance count, volume above baseline | An unambiguous session: most issues rose and a lot of stock changed hands. The summary adds nothing to the price, and saying so is honest. |
| Index up, narrow advance count | A few large members carried it. This is the configuration the whole breadth section exists to expose, and it is invisible in the headline. |
| Index up, volume weighted to the decliners | The count and the money disagree. Check for a single very heavily traded issue before drawing any conclusion. |
| Index up, both new highs and new lows elevated | A divided market, some issues at twelve-month highs while others make lows. Worth recording, and not a forecast. |
All four rows have the same headline. That is the point of the exercise: the index move is the same number in every case, and the five figures underneath it describe four different sessions.
What a summary should not do
Attribute a cause it cannot observe. "Stocks fell on rate worries" is a story attached to a number, and the same worries would have been credited with a rise. A summary can state what happened and can note a scheduled event known in advance (an expiry, a rebalance, a data release), because those are facts about the calendar rather than inferences about motive.
Quote a threshold as though it were a constant. Any level in the five (a breadth ratio, a TRIN reading, a percentage at new lows) was calibrated on some exchange in some decade. Express it as a percentile of the series you actually have, which is the instruction every breadth page in this reference ends with.
Present the same five numbers as a forecast. Each of them is a description of a session that has finished. Read across a run of sessions they describe a condition (participation narrowing, volume expanding), and a condition can persist for months while remaining an accurate description. That is the site’s general position, set out on the method page, and a daily summary is where it is easiest to forget.
Frequently asked questions
Why does this page not carry a market summary?
Because a static page cannot honestly carry one. A summary describes a specific session, and this page will be here in five years, a dated snapshot presented as current is wrong by the next morning and stays wrong indefinitely. What is here instead is the specification: which numbers a summary should contain and what each one supports, which remains true whatever the market does.
What are the five numbers?
How many issues advanced against declined; the session’s volume as a ratio to the market’s recent average; the counts of new 52-week highs and new lows, kept separate; the average percentage range per issue; and up against down volume. Between them they answer how many took part, what the move cost in stock, whether the extremes are crowded, how far the participants travelled, and whether the money agreed with the count.
Why is the index level not one of them?
Because it is the least informative number on the page and the only one everybody already knows. A capitalisation-weighted index is dominated by its largest members, so it can make a new high while most of the list falls, that gap is the entire reason the breadth measures exist. The index level belongs in a summary as context, not as the finding.
What does "volume against baseline" mean here?
Total market volume divided by its own average over the previous twenty sessions, so the answer is a multiple rather than a share count. A raw total is unreadable. It grows with listings and float over the decades and means nothing next to a figure from another era. The average volume page sets out the window choices and the traps, including the one that matters most: today should not be inside its own baseline.
Why keep new highs and new lows separate?
Because the state worth noticing is both being large at once, and a difference hides it completely. A market with many issues at twelve-month highs and many at twelve-month lows is internally divided, which no net figure can show. The new-high/new-low page also works through the twelve-month window that suppresses the high count for a year after a peak, a fact any summary quoting new highs should state.
How should the five be read together?
As a description of one session, in that order, and with the disagreements as the interesting part. A rise with a broad advance count on expanding volume is unambiguous and adds nothing to what the price said. A rise on a narrow count, or with the volume weighted to the declining side, is the configuration where the summary earns its place, and none of the five, in any combination, is a forecast.
What should a summary explicitly not do?
Attribute the session to a cause it cannot observe. "Stocks fell on inflation worries" is a story attached to a number, and the same number would have been attributed to something else had the market risen. A summary can state what happened and note a scheduled event that was known in advance — a rebalance, an expiry, a data release — which is a fact rather than an inference.
Where does the data come from?
The exchanges publish advance, decline, unchanged and up/down volume figures alongside the new-high and new-low counts, and data providers redistribute them; historical series are generally licensed rather than free. That licensing is why this reference computes its own figures from free published sources such as FINRA’s daily files and illustrates everything else with clearly labelled synthetic series.