Index up while most stocks fall: 24 of 180 sessions

The index closes green. Your screen is red. Nothing is broken on your end, and the explanation is not that you picked badly. An index is a weighted average of a small number of large companies. The market is several thousand stocks, and on a given day those two things are measuring different populations. We counted how often they point opposite ways.

1. How often the two disagree

We took 180 sessions and, on each one, counted the qualifying symbols that closed above their prior close. If more than half closed higher, the majority was up. Then we compared that against the direction of the index proxy on the same day. A median of 6,182 symbols cleared the price and turnover filter on a given session, so the majority is being read off several thousand names, not a handful.

2. It cuts both ways

The interesting part is the split. If the index were systematically flattering a market that was quietly falling apart, the disagreements would pile up on one side. They do not. Both proxies disagree in the direction that makes the index look weak slightly more often than in the direction that makes it look strong. A day when the index falls while most stocks rise is exactly as ordinary as the reverse. That matters, because the phrase you usually hear attached to this measurement is narrowing breadth, always as a warning. In this window the measurement simply does not carry that meaning.

3. The narrower the basket, the more it disagrees

QQQ tracks a narrower and more concentrated set of companies than SPY, and it disagreed with the broad market 1.7 times as often. That is the mechanism in one number. Concentration is not a small effect in US equities: half of all dollar volume sits in 93 names, and the handful at the very top are in the top ten on essentially every session. An index built on that top slice can be carried by a few names moving together while the other several thousand go the other way. The single sharpest case in the window was 2026-04-29, when QQQ closed up 0.61% and only 27.2% of qualifying stocks closed higher. Nearly three out of four names fell on a day the Nasdaq proxy finished green. On the SPY side the sharpest was 2026-07-31, up 0.72% with 44.3% of stocks higher.

4. Most sessions are not close calls

Before reading too much into a 50% threshold, it is worth knowing how far from that line a typical session sits. The middle half of sessions runs from 37.0% to 63.7%, which is a wide band. Only 34 of 180 sessions, 18.9%, landed between 45% and 55%, while 50 sessions, 27.8%, were lopsided enough to put 70% of names on one side or 70% on the other. So the majority is usually decisive. When the index disagrees with it, the index is not splitting hairs against a coin flip; it is pointing the other way from a clear majority of the market. For completeness, the majority closed up on 97 of the 180 sessions and down on 83.

5. What to do with this

Three things follow, none of them a trade. First, stop using the index as a read on your own positions. If you hold individual names, particularly outside the largest companies, the index tells you about a weighted average you do not own. On one session in 7.5 it is pointing the wrong way entirely. Second, if you want a read on the market you are actually trading, count stocks rather than watching a price. The share of names closing higher is a different instrument from the index and it is trivial to compute. Third, treat narrowing breadth headlines with the same suspicion you would treat any other single-session statistic. The disagreement runs both ways in near-equal numbers, and the direction that gets written up is the one that fits the story being told that week. Limitations SPY and QQQ are tradable products, not the indices themselves, and each is one of the several thousand symbols in the population being counted. A single name out of 6,182 does not move the breadth figure, but the proxies are not independent of the thing they are compared against. Breadth here counts every qualifying name once, regardless of size. That is the point of the measurement, but it means a large company and a small one carry the same weight, which is the opposite of how the index is built. The two numbers are meant to differ; this measures how often they differ in sign. The universe is filtered only by price and turnover, so it includes exchange-traded funds, warrants, units and preferred issues alongside common stock. A common-stock-only version would not necessarily give the same breadth. Closes that were exactly unchanged were counted as not up. On a session with many unchanged names this biases breadth slightly downward. Reverse splits show up in unadjusted daily bars as enormous one-day gains, so rows moving more than 80% against the prior close were dropped. That removed 443 rows, 0.04% of the observations, which is far too few to affect a majority count either way. The window is 180 sessions across nine months. Concentration in US equities has been increasing for years, and a longer or earlier window would very likely give a different disagreement rate. Nothing here says how this rate has changed over time.