An index tells you what the market did. Breadth tells you how many stocks agreed with it.
About the Numbers in This Tutorial
Every figure below comes from the market session of 20 July 2026, carried through the whole tutorial as a single worked example. Open the tool today and your numbers will differ, because every session adds a reading and shifts the percentiles built from the stored record. The reasoning is what transfers, not the readings.
Counting, not weighting
A cap-weighted index can rise on the back of a handful of enormous companies while most of its members fall. The index has no way of showing you that, because a $4 trillion company and a $4 billion one do not carry equal votes in it.
Breadth gives every stock exactly one vote. It counts how many closed higher than the day before and how many closed lower, and it does nothing else. That single change of method is the whole idea.
What this tool counts
The universe is roughly 3,400 US common stocks, rebuilt every session. It is common stock only, so funds, trusts and other instrument types are filtered out before anything is counted.
The exact number moves a little day to day as listings come and go, and it is shown on the page, which is worth glancing at: a session counting 3,428 names is a different measurement from one counting 3,100.
A Worked Session
On 20 July 2026 the tool counted 3,428 stocks: 1,111 advancing and 2,279 declining. Two out of every three stocks fell that day. No index level tells you that on its own.
The same twelve stocks, counted two ways. Size decides the index; a headcount decides breadth.
Why the percentages do not add to 100
That session shows 32.4% advancing and 66.5% declining, which sums to 98.9%. The missing slice is stocks that closed exactly unchanged. They are counted in the universe but belong to neither side, so they quietly go missing from the two percentages.
It is a small gap on a normal day and a useful sanity check: if the two figures nearly add to 100, almost everything moved.
The row of tiles at the top describes the most recent settled session. Each one carries a number, a description, and a line underneath that is easy to overlook and is the most valuable part.
The eight tiles
Advancers and Decliners, as counts with their share of the universe.
Daily A/D Ratio, advancers divided by decliners. Above 1 means more rose than fell.
5-Day and 10-Day A/D Ratio, the same division applied to the summed advancers and decliners over that trailing window, which smooths out a single unusual session.
New 52-week Highs and New 52-week Lows, counts of stocks reaching a one-year extreme that day.
Breadth Streak, a sentence rather than a number: how many consecutive sessions have had more advancers than decliners, or the reverse.
The percentile line is the point
Under most tiles sits something like "13th percentile of 92 sessions". That line is doing the interpretive work, because a raw count means nothing without knowing what normal looks like.
On 20 July 2026 the advancing share was 32.4%. Read alone that is just a number. Read against the stored record it sits at the 13th percentile, meaning only about one session in eight has been worse for breadth. That is the difference between "a down day" and "a notably broad down day".
Advancing share across the stored record
Value
Weakest session
16.0%
Median session
51.2%
Strongest session
84.4%
The session above (20 July 2026)
32.4%
92 stored sessions as of 20 July 2026. The median sitting near 51% is what makes the extremes readable.
Careful
A percentile is only as deep as the record behind it. "92 sessions" is a few months, not a market cycle, so the strongest session in the record is the strongest recently. The count is printed alongside for exactly this reason, and the archive grows by one every trading day.
Chapter 3
All Stocks vs the S&P 500
The toggle above the tiles switches the entire page between two universes. Everything below it recalculates.
Two very different populations
All Stocks is the full count, around 3,400 names, most of them small and mid caps. S&P 500 narrows to roughly 500 large-cap index members. The second is not a sample of the first, it is the top end of it.
Take 20 July 2026. The two universes agreed almost exactly on direction: 32.4% of all stocks advanced against 33.3% of S&P members. On the face of it, the same day.
Where they part company
They disagree completely on how violently things moved. Averaged across every session both universes cover:
All Stocks
S&P 500
Average share moving 3% or more
27.2%
15.7%
Stocks moving 3%+ on 20 July 2026 (up)
6.0%
1.2%
Stocks moving 3%+ on 20 July 2026 (down)
12.4%
3.4%
Across the 66 sessions both universes cover. On a typical day, more than a quarter of all listed stocks move 3% or more; among S&P members it is closer to one in six.
This is the size effect made countable. Smaller companies simply travel further on an ordinary day, so any magnitude reading from the All Stocks view will look dramatic next to the same reading from the index.
The practical consequence: use All Stocks when you want to know what the market as a whole did, and the S&P view when you want breadth that is comparable with the index most people quote.
The S&P history is shorter, on purpose
The S&P view carries fewer stored sessions than All Stocks, and the page says so when it does. The second universe was added later, so its archive started later and grows by one session a day like everything else. Percentiles in that view are therefore ranked against a smaller record.
The choice is a deep link, so a view can be shared exactly as it looks:
Below the tiles, two charts do different jobs. The first tracks the ratio. The second is where breadth becomes interesting.
Advance / decline ratio
The trailing A/D ratio plotted over time, with 1.0 as the dividing line. Above it more stocks have been rising than falling; below it the reverse. This is the smoothed version of the daily tile, and it answers whether a weak session was isolated or part of a run.
Cumulative breadth, and the divergence it exposes
The second chart is a running total. Every session adds its net advancers, advancers minus decliners, to a line that only turns when the balance turns. Alongside it runs a second cumulative line of new highs minus new lows, and behind both sits SPX indexed to the same starting point.
The reason all three share a chart is divergence. When the index keeps climbing while the cumulative breadth line flattens or rolls over, the advance has narrowed: fewer and fewer names are carrying it. The two lines separating is the signal the chart exists to show, and it is invisible on a price chart alone.
A bar panel underneath gives each session's net advancers on its own, so a single violent day is legible instead of being buried inside the running total.
How to Read a Divergence
Cumulative breadth is not a price and its level means nothing on its own. Only its direction relative to the index matters. Both rising together is a broad advance. The index rising while breadth falls is a narrowing one. Neither is a timing signal, and divergences can persist far longer than they look like they should.
New highs minus new lows is worth watching separately. On 20 July 2026 there were 80 new 52-week highs against 81 new lows, a near dead heat, on a day when two thirds of all stocks fell. The two measures gave slightly different readings of the same session, which is exactly why the page shows both.
Chapter 5
The Daily History Table
Every stored session, newest first, one row each.
The columns
Date, then advancers / decliners as raw counts.
A/D, that session's ratio.
Net breadth, drawn as a bar so the shape of a run is visible by scrolling rather than by reading.
Adv %, the advancing share.
5%+ Movers, how many stocks moved at least 5% in each direction. This is the magnitude column: breadth can be even while the size of the moves is not.
5D and 10D A/D, the trailing ratios as of that session.
Thrust and washout
A session where 80% or more of stocks advanced is tagged Thrust; one where 80% or more declined is tagged Washout. Crossing 90% strengthens the tag.
These are rare, which is the point of flagging them. Across the 92 stored sessions there were 2 thrust days and 3 washout days: five occasions in roughly four months where the market moved as one.
Reading Tip
Rarity is what gives these tags meaning, and it is also what makes them a poor basis for statistics. Five events is an observation, not a sample. Treat a thrust tag as something worth looking at, never as something with a measured success rate behind it.
Expanding a row
Any row opens to show the full magnitude breakdown: how many stocks moved more than 3%, 5% and 10% in each direction. The Expand all button opens every row at once, which is the fastest way to scan for the days when the tails were unusually fat.
These thresholds are cumulative, so a stock up 12% is counted in the 3%, 5% and 10% buckets alike.