Market research · 2000 to 2026
How often does the Nasdaq fall 6.7%?
Twenty-seven years, 6,752 trading days. A single-day fall of 6.7% or more has happened 23 times, every one of them inside a crash, and not once since March 2020. A pullback of that size from the one-year high has happened 41 times, about three every two years, and the one that began on 22 October 2026 is the forty-first.
Invesco QQQ Trust, the tradeable Nasdaq-100 · daily closes 3 January 2000 to 5 November 2026 · price returns, regular session
23
single days down 6.7% or more
41
pullbacks of 6.7%+ from the one-year high
1.69%
average daily move, one standard deviation
84%
of days within plus or minus 2%
Why 6.7%
Four standard deviations, forty-nine times.
The threshold is four standard deviations of a QQQ daily return. The index has moved 1.69% a day on average since 2000, and 84% of all days land within plus or minus 2%. A normal distribution would deliver a four-sigma day about once in a working lifetime. The market has delivered 49 of them in 27 years, 23 down and 26 up. The tails are fat, and they are symmetric.
Trading days by size of daily move, one-percent bins. Dark bars are ordinary days, red bars are falls of 7% or more, gold bars rises of 7% or more. The vertical scale is logarithmic so the tails can be seen at all.
Single-day falls
A 6.7% day only ever comes inside a crash.
Three quarters of the large down days belong to 2000 and 2001. The rest are the autumn of 2008 and the three weeks of March 2020. Outside those three episodes, including the 2022 bear market in which the index lost a third of its value, no single session has reached the threshold. And no episode produced just one: every crash delivered several, down and up, within weeks of each other.
| Episode | Days of −6.7% or worse | Worst day | Days of +6.7% or better |
|---|---|---|---|
| 2000 to 2001, the first internet boom unwinds | 17 | 14 April 2000, −8.7% | 21 |
| 2008, the credit crisis | 3 | 15 October 2008, −9.0% | 4 |
| 2020, the pandemic | 3 | 16 March 2020, −12.0% | 1 |
| Every other year, 2002 to 2026 | 0 | 0 |
Pullbacks
A 6.7% pullback arrives every eight months.
A pullback is counted once, from the first close more than 6.7% below the highest close of the previous twelve months until a new one-year high is set. By that count there have been 41 since 2000, in bull markets and bear markets alike. The whole of 2022 counts as a single episode, because the index never regained its one-year high until the bear market was over.
| The last five | From | To | Depth |
|---|---|---|---|
| Tariff shock | 27 Feb 2025 | 24 Jun 2025 | −22.9% |
| Winter 2025-26 | 20 Nov 2025 | 15 Apr 2026 | −12.2% |
| Summer 2026 | 10 Jun 2026 | 22 Sep 2026 | −11.3% |
| Now | 22 Oct 2026 | ongoing | −14.8% so far |
Left: pullbacks of more than 6.7% from the one-year high, by the year they began. Right: the most recent, with the depth at the low.
What we take from it
Routine pullbacks, rare days.
One
A 6.7% pullback is routine
It arrives about three times every two years, in good markets and bad, and on its own says little about what follows. The current one is the third in twelve months.
Two
A 6.7% day is not
It has only ever appeared inside a crash, and never alone. Every episode produced several, which is the signature of a market that has stopped functioning normally.
Three
The tails are symmetric
The 26 best days of the period were as extreme as the 23 worst, and most came inside the same weeks. Being out of the market on the bad days has usually meant being out on the good ones.
Four
Volatility clusters
Three clusters account for 23 of 23 large falls. A risk model that treats each day as independent underestimates both how often the losses come and how tightly they bunch.
How we built it
QQQ daily closing prices from Interactive Brokers, 3 January 2000 to 5 November 2026, split-adjusted, regular session, returns close to close and price only. Single-day falls are counted at 6.7% or worse. A pullback is counted once, from the first close more than 6.7% below the highest close of the previous 252 trading days until a new one-year high; counting from the all-time high instead gives 21 episodes, because the March 2000 high was not regained until 2016. The standard deviation is the simple daily figure for the whole period.
Important information
This page is for general information and discussion only and is not investment, legal or tax advice, nor an offer or recommendation to buy, hold or sell any security. Historical frequencies are not forecasts; the next large fall need not resemble the last. All investments involve risk, including loss of capital. Past performance is not a reliable indicator of future results. Analysis as of 6 November 2026.