RiskUpdated daily after close · as of 2026-09-04
Return Dispersion: How Differently the 100 Largest Stocks Are Moving
Each week we measure the spread of returns across the 100 biggest U.S.-listed stocks — the cross-sectional standard deviation, back to 2011. A wide spread under a small index move means single stocks are living through something the index never shows: the moves are violent, opposite, and cancelling inside the average. This is the realized-market twin of the options market's dispersion pricing.
This week's reading
For the week ending 2026-09-04, the return spread across the top 100 was 4.04 percentage points — the 76th percentile of all 818 weeks since 2011, rank 195 widest ever measured this way. The average stock in the basket moved 2.98% in absolute terms while SPY itself moved +0.11%. State: Elevated.
Sources, methodology & freshnessLast updated 2026-09-04 · Open ↓Close ↑
Source
Computed from our own ~5,500-symbol daily price database (TradeStation), with the top-100 ranking from the Polygon share-class snapshot (cap recomputed daily as shares × latest close) and SPY as the index reference.
Methodology
Weekly returns from the last trading day of each ISO week; cross-sectional population standard deviation across the basket (minimum 60 names priced); regime states at fixed percentiles (50/80/95) of the full 2011+ history; forward returns are SPY averages from each week-end, with 21/63-session windows overlapping across adjacent weeks.
Updates
Recomputed after every US trading close as part of the daily pipeline.Last: 2026-09-04
Maintained & reviewed by Yuriy Matso — methodology shown on the page.
Return dispersionweek of 2026-09-04
4.04pp
Elevated — 76th percentile since 2011
Avg |move|
2.98%
SPY week
+0.11%
Rank
#195/818
Cross-sectional standard deviation of the basket's weekly returns (100 names priced this week). Regime bands sit at fixed percentiles (50/80/95) of the spread's own history: 3.3 / 4.2 / 5.8pp.
01
The weekly spread since 2011
The gauge itself, with SPY above for the context that decides everything. The 2020 spikes came with an index crash — stocks moving a lot, together. The recent extremes are the other kind: crisis-grade spread under an index near highs, single-stock moves cancelling before they reach the surface.
Range:
02
Who paid for it — this week's widest movers
The spread is one number; these are the names that produced it. One glance tells you whether the dispersion is a sector event or a market-wide repricing.
Biggest gainers
- SNDK+17.17%
- DELL+14.98%
- DE+10.03%
- MU+9.17%
- INTC+7.07%
Biggest losers
- PANW-10.32%
- PLTR-6.42%
- UNP-5.79%
- SCCO-5.26%
- RTX-5.16%
03
What happened next — SPY after each dispersion regime
The honest test, run before shipping — and the split is the finding. Wide weeks overall preceded above-baseline returns, because most of them sat inside selloffs that mean-reverted. Split the extreme weeks by what the index was doing and the story divides: extreme spread inside a moving index was a washout marker (strongly above baseline), while extreme spread under a quiet index preceded below-baseline returns — on a sample small enough that we print it in bold and treat it as a base rate. It is not a trade signal.
| Week's spread | Weeks | Next 5 sessions | Next 21 sessions | Next 63 sessions |
|---|---|---|---|---|
| Lowest quartile (tight pack) | 208 | +0.14% · 60% win | +0.77% · 68% win | +2.04% · 75% win |
| Middle half | 404 | +0.21% · 56% win | +0.89% · 67% win | +2.82% · 73% win |
| Highest quartile (wide spread)This week | 206 | +0.44% · 61% win | +1.83% · 69% win | +4.74% · 81% win |
| Extreme dispersion, quiet index (|SPY| < 2%) | 18 | -0.42% · 39% win | +1.08% · 56% win | +2.50% · 73% win |
| Extreme dispersion, index moving (|SPY| ≥ 2%) | 24 | +0.47% · 58% win | +3.80% · 78% win | +9.68% · 96% win |
| All weeks | 818 | +0.25% · 58% win | +1.09% · 68% win | +3.08% · 75% win |
Average SPY price return and share of positive outcomes from each week's end, 2011+. The 21/63-session windows overlap across adjacent weeks, so effective sample sizes are smaller than the counts suggest. Dividends excluded.
04
What this gauge cannot tell you
Three limits, stated plainly. The basket is today's: the 100 names are ranked by current market cap and applied across history, so the series describes the stocks you own now. It is no point-in-time index — companies that fell out of the top 100 are absent from the past, and young names drop out of early weeks. It is direction-blind: a wide spread says nothing by itself about where the index goes — the outcome table above is the only directional claim here, and its most interesting cell is its smallest. It is realized, not predictive: for what the market expects dispersion to be, read the implied side — Volatility Premium (VIXEQ−VIX and implied correlation) — and the factor view on Momentum Churn.
Return dispersionweek of 2026-09-04
4.04pp
Elevated — 76th percentile since 2011
Avg |move|
2.98%
SPY week
+0.11%
Rank
#195/818
Cross-sectional standard deviation of the basket's weekly returns (100 names priced this week). Regime bands sit at fixed percentiles (50/80/95) of the spread's own history: 3.3 / 4.2 / 5.8pp.
How Return Dispersion Works
- 1Take today's 100 largest stocksThe basket is the 100 biggest U.S.-listed companies by market cap, recomputed daily as shares outstanding × latest close — the same ranking our market heatmap uses. It is today's list applied across history, and the page says so plainly (that makes the history descriptive, not survivorship-clean).
- 2Compute each week's return spreadFor every calendar week since 2011, each stock's Friday-to-Friday return is computed, and the gauge is the cross-sectional standard deviation across the basket — how widely the pack disagreed that week. The average absolute move and SPY's own weekly return are stored alongside.
- 3Read it against its own historyRegime states (Tight pack / Elevated / Wide spread / Extreme dispersion) sit at fixed percentiles (50/80/95) of the full 2011+ history, so a headline like "98th percentile" means exactly that: wider than 98% of all weeks measured the same way.
- 4Check the index contextThe same spread means different things depending on what the index did. The page splits extreme weeks into "index quiet" (|SPY| under 2%) and "index moving" — historically, the quiet kind preceded below-baseline returns and the moving kind marked washout bottoms.
Who Uses Return Dispersion
Stock pickers
Dispersion is the size of the penalty for owning the wrong names. A 98th-percentile week means single-stock risk is at crisis levels even when the index chart looks serene.
Options traders
The realized-market check on the dispersion trade: compare this spread against what the options market is pricing on the Volatility Premium page (VIXEQ−VIX and implied correlation).
Risk managers
Index-hedged books are unhedged against dispersion. When the spread is extreme and the index is quiet, portfolio tracking error explodes while VaR models see nothing.
Market watchers
The week-by-week record of whether "the market" is one trade or a hundred separate ones — and the honest base rates for what followed each regime.
Pro Tips
01
The index context is the signal
Extreme dispersion inside a crashing index has historically been a washout marker — strongly above-baseline forward returns. Extreme dispersion under a QUIET index is the rarer and weaker configuration: below-baseline short-horizon returns in our history, on an admittedly small sample.
02
Watch the clustering
Dispersion regimes persist. One wide week is earnings season; a cluster of top-percentile weeks across months — like 2026 has produced — is a regime, and it usually shows up in the implied-correlation market too.
03
Pair it with the movers
The spread is one number; the movers table under it shows who paid it. A dispersion spike led by one sector (say, semiconductors) reads differently from one spread across the whole basket.
04
Don't compare levels across eras
Mega-cap composition changes: today's basket is tech-heavier than 2011's. Percentiles against the same construction are the honest read; the raw pp level is secondary.
Common Issues & Solutions
Is this survivorship-biased?▾
The basket is today's 100 largest names applied backward, so yes — deliberately, and it is labeled as such. Companies that shrank out of the top 100 are missing from history, and recent IPOs/spinoffs drop out of early weeks (each week reports how many names it averaged). The alternative — historical point-in-time membership — has no free daily source. Same trade-off, same disclosure, as our Bear Market Breadth tool.
Why the standard deviation and not the average move?▾
The average absolute move (also shown) blends direction out. The cross-sectional standard deviation measures disagreement — a week where everything falls 3% has a big average move but low dispersion; a week where half rise 5% and half fall 5% is pure dispersion. The second kind is what an index hides.
The forward returns look bullish after wide weeks — why is the current reading flagged?▾
Because the bullish history comes almost entirely from wide weeks INSIDE selloffs — crash washouts that mean-reverted. Wide weeks under a quiet index (the current kind, when flagged) are the below-baseline subset. The outcome table shows both splits separately.
Why weekly rather than daily?▾
Daily cross-sections are dominated by single-name earnings noise. A week is long enough for a rotation to register and short enough to catch regime changes — and it matches how the dispersion literature (and the options market's weekly cycle) frames the question.
Frequently Asked Questions
What is return dispersion?▾
The spread of returns across stocks over the same period — here, the cross-sectional standard deviation of weekly returns across the 100 largest U.S.-listed stocks. High dispersion means stocks are moving very differently from one another; low dispersion means the market is trading as one block.
Why can dispersion be extreme while the VIX is low?▾
Index volatility is roughly average stock volatility scaled by the correlation between stocks. When correlation is very low, violent single-stock moves cancel inside the index — so the VIX can sit in the teens while individual stocks swing like a crisis. The Volatility Premium page shows the options market pricing exactly that.
Is high dispersion bearish?▾
Not by itself — the context decides. In our 2011+ history, extreme-dispersion weeks inside a moving index (crashes) preceded strongly above-baseline returns, while extreme dispersion under a quiet index preceded modestly below-baseline returns on a small sample. It is a regime gauge first, a directional signal barely.
What counts as extreme?▾
The 95th percentile of the full 2011+ weekly history — about a 5.75 percentage-point spread. The regime bands (50/80/95) are fixed percentiles of the series' own history, not fitted to outcomes.
How is this different from implied correlation (COR3M)?▾
COR3M is what options prices expect stock-to-stock correlation to be over the next three months; this tool measures what dispersion actually was, week by week. Reading them together — priced dispersion vs realized dispersion — is the point of publishing both.
Explore Other Tools
Volatility Premium
What the options market charges for this: VIXEQ−VIX and implied correlation, the forward-looking twin
Momentum Churn
The factor view of the same storm — winners-minus-losers volatility against a quiet index
Market Heatmap
The same top-100 basket as a treemap — who is actually moving today
Hidden Bear Index
The average stock vs the index — dispersion's slower cousin, in drawdown terms
Last updated: 2026-09-04