Key points
- The Nasdaq Composite closed lower for a sixth straight session on July 29, down 5.4% over the run and 9.8% below its June 2 record.
- It has happened 85 times since February 1971. This is the first six-day streak in 569 sessions, since April 2024.
- One month out, the Nasdaq gained a median 1.5% after these streaks. The median for any random month is also 1.5%.
- Three months out it did worse than average: +2.7% against a +3.6% baseline.
- Every streak in 2000, 2001 and 2002 was followed by losses of 15% to 46% a year later. Every streak after 2009 was followed by gains.
- The index understates the damage. July cost Sandisk 55.3%, KLA 43.7% and Micron 36.1%.
The Nasdaq Composite closed lower on Wednesday for the sixth day in a row. By Thursday morning there will be a dozen articles telling you what that means.
So we pulled every Nasdaq session since February 1971 and counted. The honest answer is that a six-day losing streak tells you almost nothing, and the number that proves it is the one these articles usually leave out.
What the streak actually looks like
The index finished at 24,442.94, which is 5.4% below the 25,837.21 it closed at on July 21. Two sessions did most of the work and three of them barely moved.
| Session | Close | Change |
|---|---|---|
| July 22 | 25,690.90 | -0.57% |
| July 23 | 25,137.69 | -2.15% |
| July 24 | 24,975.82 | -0.64% |
| July 27 | 24,932.08 | -0.18% |
| July 28 | 24,876.91 | -0.22% |
| July 29 | 24,442.94 | -1.74% |
The Nasdaq set a record close of 27,093.90 on June 2 and sits 9.8% under it. July has cost 6.8%. The index is still up 5.2% for 2026.
How often does the Nasdaq fall six days in a row?
Eighty-five times before this one, across 13,986 sessions going back to February 5, 1971. That works out to about once every 20 months.
The last one ended on April 19, 2024, which is 569 sessions ago. Streaks this long are not rare, but they are not common either, and going more than two years without one is the unusual part of this setup.
Of those 85 streaks, 42 stopped at exactly six days. Twenty-two ran to seven, thirteen reached eight, four got to nine and four ran past ten. The record is 16 straight down closes, ending February 9, 1984.
So roughly half the time, six is where it stops.
What happened next?
Here is the table almost nobody prints, because the interesting column is the second one. Median forward returns from the last close of a six-day streak, against the median for any randomly chosen starting point in the same 55 years.
| Horizon | After a 6-day streak | Any random day | Positive after streak | Positive normally |
|---|---|---|---|---|
| 1 week | +0.8% | +0.4% | 66% | 58% |
| 1 month | +1.5% | +1.5% | 60% | 61% |
| 3 months | +2.7% | +3.6% | 60% | 66% |
| 6 months | +7.1% | +7.0% | 67% | 70% |
| 12 months | +18.3% | +15.0% | 76% | 76% |
There is a real edge at one week. Two thirds of these streaks were followed by a higher index five sessions later, against 58% for a normal week. That is a bounce, and it is the only place in the table where the signal beats the coin.
After that it disappears. At one month the streak result and the baseline are identical. At three months the streak did worse than a random entry, and at six months it was a wash. At twelve months the median is better but the odds of being up at all are exactly the same 76%.
A six-day losing streak is not a buy signal and it is not a warning. It is mostly just six down days.
What the average is hiding
The averages are useless here, and the individual outcomes are the reason. Here is every six-day streak since 2000, all 26 of them, with what the index did afterward.
| Streak ended | Days | Streak drop | 1 month later | 3 months later | 12 months later |
|---|---|---|---|---|---|
| Oct 12, 2000 | 6 | -12.7% | -1.5% | -14.6% | -46.2% |
| Nov 13, 2000 | 6 | -14.0% | -4.9% | -16.0% | -34.8% |
| Dec 20, 2000 | 7 | -22.6% | +21.8% | -17.3% | -15.3% |
| Jun 18, 2001 | 7 | -12.2% | +1.4% | -28.4% | -26.6% |
| Aug 10, 2001 | 6 | -6.3% | -19.3% | -2.7% | -30.4% |
| Apr 24, 2002 | 6 | -5.7% | -0.9% | -24.7% | -14.9% |
| Aug 8, 2003 | 6 | -5.2% | +14.0% | +20.2% | +10.0% |
| Jan 7, 2005 | 6 | -4.1% | -0.1% | -4.6% | +11.0% |
| Mar 9, 2006 | 6 | -2.8% | +4.0% | -4.6% | +6.8% |
| May 18, 2006 | 8 | -7.0% | -3.2% | -1.0% | +18.3% |
| Jun 13, 2006 | 8 | -6.6% | -0.9% | +6.9% | +25.4% |
| Aug 16, 2007 | 6 | -6.2% | +5.3% | +7.9% | +0.1% |
| Jan 8, 2008 | 8 | -10.4% | -6.0% | -4.9% | -34.5% |
| Oct 9, 2008 | 7 | -21.4% | +0.1% | -4.5% | +30.0% |
| Feb 23, 2009 | 6 | -10.0% | +9.3% | +21.9% | +59.5% |
| Nov 25, 2011 | 7 | -9.1% | +7.5% | +22.3% | +22.5% |
| Oct 12, 2012 | 6 | -3.3% | -6.5% | +2.4% | +26.1% |
| Aug 25, 2015 | 6 | -11.5% | +5.1% | +13.2% | +15.8% |
| Sep 29, 2015 | 6 | -6.5% | +12.8% | +13.1% | +17.7% |
| Jan 11, 2016 | 8 | -9.2% | -7.6% | +5.0% | +19.7% |
| Apr 29, 2016 | 7 | -3.5% | +3.6% | +8.1% | +27.6% |
| Nov 4, 2016 | 9 | -5.0% | +5.7% | +12.4% | +34.5% |
| Aug 5, 2019 | 6 | -7.3% | +3.2% | +8.5% | +41.6% |
| Sep 6, 2022 | 7 | -8.7% | -3.4% | -2.6% | +19.1% |
| Oct 12, 2022 | 6 | -6.8% | +6.7% | +5.6% | +28.7% |
| Apr 19, 2024 | 6 | -7.1% | +9.9% | +17.8% | +9.3% |
Every streak in 2000, 2001 and 2002 was followed by a double-digit loss a year later, and three of them by losses worse than 26%. The January 2008 streak was followed by a 34.5% decline. Every one of the twelve streaks from 2009 onward was followed by a gain twelve months later.
The middle of the table is the honest part. Between 2003 and 2007 the outcomes were mildly positive and unremarkable, and the August 2007 streak was followed by a 0.1% gain a year later, which is as close to nothing as this dataset produces. The signal was not warning anyone about what 2008 would do.
The streak was not the variable. The regime was. In an unwinding bubble, six down days was an early symptom. In a bull market, it was a dip that resolved higher. The count itself cannot tell you which of those you are standing in, which is exactly the problem with using it.
What is driving this one
Memory and chip equipment. Micron (MU) fell 10.1% on Wednesday and KLA Corporation (KLAC) fell 11.0%, part of a chip selloff that has run for weeks. SK Hynix guided 2026 capital spending 50% higher, to at least $31 billion, and CXMT went public in China on a 466% debut, putting a second large memory supplier in front of investors.
The index number badly understates what happened underneath it. Measured from the June 30 close through Wednesday, this was a bear market in the chip complex, not a pullback.
| Stock | June 30 close | July 29 close | July |
|---|---|---|---|
| Sandisk (SNDK) | $2,273.73 | $1,016.05 | -55.3% |
| KLA (KLAC) | $301.71 | $169.82 | -43.7% |
| Applied Materials (AMAT) | $723.00 | $437.08 | -39.5% |
| Micron (MU) | $1,154.29 | $737.88 | -36.1% |
| Nvidia (NVDA) | $200.09 | $190.06 | -5.0% |
Sandisk lost more than half its value in a single month. Nvidia lost 5.0%. The Nasdaq lost 6.8%. A cap-weighted index built on the largest companies will always report the mildest version of a selloff that is concentrated further down the stack.
The Federal Reserve held rates on Wednesday in a 9 to 3 vote, and chairman Kevin Warsh named the AI buildout as one of four questions the committee argued over. "The business capex boom, for example, is driving up prices of memory and logic chips and associated A.I. infrastructure," Warsh said, before asking whether that indicates a broader inflationary problem.
The companies reporting into the selloff have said the opposite of what the tape is saying. Lam Research (LRCX) closed down 7.0% Wednesday and then posted record revenue, margin and earnings after the bell.
"Lam delivered record revenue, operating margin and earnings per share in the June quarter as AI-driven demand continues to reshape the semiconductor industry," said Tim Archer, the company's president and chief executive officer.
Dave Mosley, chief executive of Seagate (STX), told analysts this week that the vast majority of nearline exabytes are already committed into calendar 2028 under long-term agreements, and that customers are pushing their planning horizons further out.
That gap is what makes this streak hard to read. Traders are selling an AI spending peak that the companies keep saying has not arrived.
Seven straight would put this in the top 30 streaks since 1971. About half of them stopped at six.
Nasdaq Composite history is computed from daily closes for the index from February 5, 1971 through July 29, 2026, sourced from Yahoo Finance. A streak is a run of consecutive sessions closing below the prior session. Forward returns are measured from the final close of each streak at 5, 21, 63, 126 and 252 trading sessions. Baseline figures use every possible starting session in the same period. Individual stock prices are last regular-session trades on June 30 and July 29, 2026. This is general market commentary and not investment advice.
