Skip to Content

UPDATES

Tail Risk with Volatility Compression

Follow-up · Nasdaq-100

How Rare Is a Month Inside One Day’s Range?

Dropping the 10% rally filter: 30 years of 23-day Nasdaq windows versus the prior session

In our previous blog post we discussed how rare it was to see a strong 5-day rally of nearly 10% followed by the market spending the next 23 days mostly in the range of the last bar of the rally. Over the next 23 days after August 4, 56% of the daily ranges were within the August 4th range. There are about 23 trading days in one month, and in this case that window is August 5 through September 4. In 30 years, we have never seen this level of contraction.

After thinking about it, I asked the question: how rare is it, during any 23-day period, for the market to spend most of its time within the range of the previous one day — the single session immediately prior to those 23 days — regardless of what kind of move came before it?

These are the results.

The test

Take every Nasdaq-100 cash session from September 5, 1996 through September 4, 2026. Call that session day 0. Then look at the next 23 trading days and count how many of those 23 printed fully inside day 0’s high and low — the entire daily range, high and low, nested inside yesterday’s bar.

August 4, 2026 scored 13 of 23, or 56.5%. That is the bar we are measuring against. No 5-day rally filter. No VIX filter. Just: after any given day, did the following month live inside that one bar?

There were 7,525 such windows in the market over 30 years.

The headline

In 30 years of Nasdaq-100 cash, only 5 sessions — 0.07% of all days — were followed by 13 or more of the next 23 sessions trading entirely inside that day’s range. The median day contains zero of the next 23. August 4 is tied for third-highest containment in the entire sample.
How many of the next 23 days sit inside yesterday’s bar Count of day-0 sessions Share of 30 years
≥ 1 day (at least one fully inside) 2,592 34.5%
≥ 5 days 276 3.7%
≥ 8 days 77 1.0%
≥ 10 days 35 0.47%
≥ 12 days 16 0.21%
≥ 13 days (56.5% — this month) 5 0.07%
≥ 17 days 2 0.03%
All 23 days 0 0.00%

About two-thirds of all sessions (4,933 of 7,525, or 65.6%) are followed by a month in which not a single day stays inside that bar. The 90th percentile is 2 days. The 99th percentile is 8 days. The mean is 0.78 days. Getting to 13 is the extreme tail.

Next-23-day inside-bar count (Nasdaq-100 cash, 1996–2026)

0

4,933
1

1,286
2

577
3

282
4

171
5

104
6–9

137
10–12

30
13

3
17

1
20

1

Bar width is relative to the 4,933 sessions with zero follow-through days inside. Orange marks the 13+ tail that includes August 4, 2026. Source: NDX cash daily, 1996-09-05 to 2026-09-04.

The only five times it happened

These are every day-0 session in 30 years that contained 13 or more of the next 23 daily ranges. August 4 is on the list — and it is the narrowest container bar of the five by a wide margin.

Day 0 (the container bar) Next 23 days inside Day-0 range What that day was
2010-05-06 20 / 23 11.40% Flash crash
1997-10-27 17 / 23 8.02% Asian-crisis crash day
2017-06-09 13 / 23 4.15% Sharp tech selloff
2020-09-04 13 / 23 6.09% Post–Labor Day dump
2026-08-04 13 / 23 2.45% This episode

The other four were panic days. Their ranges were 4% to 11% of the index. If you print an 11% flash-crash bar, of course the following month can sit inside it — the container is enormous.

August 4 was only 2.45% wide. A typical Nasdaq day in this sample has a 1.46% range. The 75th percentile is 2.28%. So August 4 was a large day, but not a crisis day. It is the only time in 30 years that the next month lived inside a bar of roughly normal size.

The distinction matters. “56% of the next 23 days inside yesterday” has happened four other times — always because yesterday was a crash range. It has happened once after a 2.45% bar, and that once is August 4, 2026.

What this adds to the first post

The original finding still stands: a ~10% five-day climactic rally, then this kind of sleep, has no match in 30 years.

The follow-up question was whether the 56% inside-bar statistic was just an artifact of staring at a big rally. It isn’t. Even if you ignore the rally completely and ask only “how often does a month live inside the prior day,” the answer is five times since 1996, and four of those five are unusable as analogs because the prior day was a 4–11% panic bar.

What makes this more significant is the tail risk created by extreme volatility compression. The Nasdaq has spent nearly a month absorbing what would normally be multiple daily ranges inside a single, non-crisis session. That does not tell us which direction the eventual break will occur, but it does tell us that the current state is historically abnormal.

Volatility does not remain compressed indefinitely. When a market becomes this tightly contained after a nearly 10% impulse, the larger risk may no longer be the quiet itself — it may be the magnitude of the move when that compression finally resolves.


Methodology

Data: Nasdaq-100 cash session OHLC, September 5, 1996 – September 4, 2026 (7,548 sessions; 7,525 windows with 23 days of follow-through). For each session i, count how many of sessions i+1 through i+23 have high ≤ high[i] and low ≥ low[i]. August 4, 2026 scored 13/23. Range percent is (high − low) / midpoint of that session. No impulse, VIX, or volume filter on this scan.

Hypothetical historical analysis has limitations. Past patterns do not guarantee future range. Cash regular-hours bars can differ from NQ Globex extent overnight. Nothing here is a recommendation to buy or sell any security or futures contract.