The Fed Crushed Stocks Wednesday. By Friday It Never Happened. We Tested All 24 Since 1952.
On July 29 the S&P 500 fell 1.52% on the Fed decision. Two sessions later the entire loss was gone and the index closed the week above where it started. That exact shape — a midweek shock, fully repaired by Friday — has happened 24 times in 74 years, and the bullish statistic those 24 cases produce is the kind that gets quoted all weekend. It is also completely ordinary, and we can show you exactly why.
The Conclusion
The 58% Monday win rate is real, and it is not an edge. Monday closed higher after 14 of the 24 precedents. But every Friday in the sample was green by construction — a green Friday is what erases the loss — and any Friday closing up 1.5%+ is followed by a 57.2% Monday. The setup adds 1.1 points to a base rate nobody quotes.
The control erases the signal. 58.3% for the setup, 57.2% for any strong Friday, 49.8% for Mondays in general. Against the matched control the setup yields a binomial p = 0.57 and a bootstrap p = 0.14. What looks like a Fed-reversal effect is end-of-week momentum wearing a costume.
The record flipped in 2010. From 1990–2009 the pattern went 8-for-9 on Monday (88.9%). Since 2010 it has gone 2-for-9 (22.2%). The famous 58% is carried almost entirely by a two-decade window that closed sixteen years ago.
“Fed day” is doing no work. Only 2 of the 24 landed on an actual FOMC decision day — November 12, 1997 (Monday +1.92%) and January 31, 2024 (Monday −0.32%). One up, one down. Seven predate 1994, when the Fed didn't announce on meeting day at all.
The V-shape is the weakest resolution, not the strongest. Over the next 20 sessions the full-recovery group averaged −0.68%. Weeks that only partly recovered averaged +1.60%, and weeks that never recovered at all averaged +1.83%. A two-day vertical repair spends demand that slower basing holds in reserve.
Everything below is how we got there — the screen, the base rates, the era split, and an interactive table of all 24 precedents where you can open the chart for every one and check us.
What Happened
The Fed decision on Wednesday, July 29 knocked 1.52% off the S&P 500. Thursday and Friday did not merely repair it — they closed the week 0.82% above the pre-Fed Tuesday close, a 2.37% round trip off Wednesday's close, with Friday finishing within 0.30% of its own high.
| Session | Close | Change | Note |
|---|---|---|---|
| Tue 28 Jul | 7,428.78 | — | pre-Fed reference |
| Wed 29 Jul | 7,316.15 | −1.52% | FOMC decision · low 7,313.92 |
| Thu 30 Jul | 7,437.63 | +1.66% | loss erased |
| Fri 31 Jul | 7,489.72 | +0.70% | closed 0.30% off its high |
A tape that rejects a Fed selloff that completely reads as strength. So we asked the only question that settles it: across every comparable week since 1952, what actually happened next — and how does that compare to what happens after an ordinary strong week?
The Test
We screened all 18,766 S&P 500 sessions from February 1952 to July 2026 for three conditions: a Wednesday closing −1.5% or worse, with the preceding session being that week's Tuesday; a Thursday and Friday following on consecutive calendar days; and a Friday close at or above the Tuesday close — the entire Wednesday loss erased. That produces 24 occurrences. Then the step that decides everything: how do Mondays behave after any strong Friday?
14 / 24
Mondays that closed higher after the setup (58.3%)
57.2%
Mondays that close higher after ANY Friday ≥ +1.5%
| Condition | n | Monday up | Mean |
|---|---|---|---|
| All Mondays | 3,463 | 49.8% | −0.04% |
| After any Friday ≥ +0.5% | 890 | 58.0% | +0.16% |
| After any Friday ≥ +1.5% | 159 | 57.2% | +0.22% |
| After this setup | 24 | 58.3% | +0.39% |
The base rate: a strong Friday already buys you a 57% Monday
Every Friday in the 24-event sample was green — it has to be, since a green Friday is what erases the Wednesday loss. Mean Friday gain: +1.77%. And any Friday like that is followed by a 57–58% Monday whether or not a Fed plunge preceded it. Against that matched pool, 14-for-24 gives p = 0.57 on a binomial test and p = 0.14 on a bootstrap of the mean. Neither is close to significant.
The Era Split Nobody Quotes
Break the 24 into thirds and the bullish case gets worse. The pattern's entire reputation was built in one two-decade window.
| Era | n | Monday up | Mean |
|---|---|---|---|
| 1952–1989 | 6 | 66.7% | +0.33% |
| 1990–2009 | 9 | 88.9% | +0.88% |
| 2010–2026 | 9 | 22.2% | −0.06% |
From 1990 to 2009 this went 8-for-9. Since 2010 it has gone 2-for-9, and across the ten most recent occurrences Monday was up just 3 times, median −0.29%. Nine observations per bucket cannot carry much weight — but anyone quoting “58% of the time stocks rise” should know where that number comes from. The two green Mondays since 2010 were August 15, 2011 and June 15, 2026; the seven red ones were all small, −0.24% to −1.79%.
Check Every Precedent Yourself
Don't take the aggregates on faith. Below is every week since 1952 that matched the screen, with the Wednesday plunge, the two recovery days, and what followed — switch the last column between the next Monday, the next five sessions and the next twenty. Click any Wednesday to open the S&P 500 chart around it. You'll find September 2008 in there, and you'll find plenty of weeks where the V-shape led precisely nowhere.
24
Precedents shown
14 (58.3%)
Higher over next monday
57.2%
Any strong-Friday base rate
Every week since 1952 where the S&P 500 fell 1.5%+ on a Wednesday and had erased the entire loss by Friday's close. The last column is what happened next, over the horizon you select. Click any Wednesday to open the chart around it and see the week yourself. † marks an actual FOMC decision day; ‡ the session right after a Tuesday decision.
| Wednesday | Wed | Thu | Fri | Fri vs Tue | Next Monday |
|---|---|---|---|---|---|
| Jun 10, 2026 | -1.62% | +1.75% | +0.50% | +0.61% | +1.65% |
| Jan 31, 2024† | -1.61% | +1.25% | +1.07% | +0.68% | -0.32% |
| Nov 9, 2022 | -2.08% | +5.54% | +0.92% | +4.31% | -0.89% |
| May 11, 2022 | -1.65% | -0.13% | +2.39% | +0.57% | -0.39% |
| Feb 23, 2022 | -1.84% | +1.50% | +2.24% | +1.86% | -0.24% |
| May 12, 2021 | -2.14% | +1.22% | +1.49% | +0.52% | -0.25% |
| Apr 15, 2020 | -2.20% | +0.58% | +2.68% | +1.00% | -1.79% |
| Oct 2, 2019 | -1.79% | +0.80% | +1.42% | +0.40% | -0.45% |
| Aug 10, 2011‡ | -4.42% | +4.63% | +0.53% | +0.54% | +2.18% |
| May 27, 2009 | -1.90% | +1.54% | +1.36% | +0.97% | +2.58% |
| Sep 17, 2008‡ | -4.71% | +4.33% | +4.03% | +3.42% | -3.82% |
| Jun 11, 2008 | -1.69% | +0.33% | +1.50% | +0.12% | +0.01% |
| Oct 16, 2002 | -2.41% | +2.23% | +0.59% | +0.35% | +1.73% |
| Oct 9, 2002 | -2.47% | +3.50% | +3.90% | +4.88% | +0.73% |
| Dec 20, 2000‡ | -3.13% | +0.80% | +2.44% | +0.03% | +0.71% |
| May 10, 2000 | -2.06% | +1.79% | +0.93% | +0.63% | +2.21% |
| Nov 12, 1997† | -1.93% | +1.18% | +1.28% | +0.49% | +1.92% |
| Nov 7, 1990 | -1.80% | +0.52% | +1.99% | +0.68% | +1.83% |
| Dec 15, 1982 | -1.57% | +0.04% | +1.63% | +0.07% | -0.90% |
| Oct 16, 1974 | -1.55% | +1.19% | +1.56% | +1.18% | +1.69% |
| Sep 4, 1974 | -2.60% | +3.17% | +0.78% | +1.28% | -2.38% |
| Jul 10, 1974 | -1.83% | -0.13% | +4.08% | +2.05% | +0.76% |
| Dec 5, 1973 | -1.53% | +2.45% | +2.21% | +3.12% | +1.49% |
| Oct 31, 1956 | -1.70% | +2.06% | +0.99% | +1.32% | +1.32% |
What Monday Actually Looks Like
Using the 23 precedents since 1962, the earliest date with real intraday range in the dataset: 83% of these Mondays traded below Friday's close at some point, and 83% also traded above it. The median worst drawdown was −0.82%; the median best extension +0.76%.
In other words the typical Monday after this setup is a two-sided chop of roughly 1.5% range that ends near where it started. An intraday dip below Friday's close is the norm, not a signal — it happened in more than four of five cases, including most of the Mondays that closed green.
The opening gap carries nothing either. On the 18 precedents with a usable opening print, the median gap was −0.06% and the open-to-close move was up only half the time.
The More Interesting Finding: The Week After
Monday is a coin flip. The following week is where this setup genuinely separates from the pack — in the wrong direction.
| From the Friday close | Up | Mean | Median |
|---|---|---|---|
| +1 day | 58.3% | +0.39% | +0.72% |
| +2 days | 54.2% | +0.11% | +0.47% |
| +3 days | 50.0% | −0.33% | −0.05% |
| +4 days | 41.7% | −0.55% | −0.41% |
| +5 days | 41.7% | −0.70% | −0.84% |
| +10 days | 50.0% | −0.37% | −0.02% |
| +20 days | 50.0% | −0.68% | +0.74% |
Against an all-sessions baseline of 56.7% up and +0.18% mean over five days, a −0.70% mean is a meaningful gap (p = 0.028 on a bootstrap against all sessions). The honest caveat: against the matched control of big up-Fridays it weakens to p = 0.149. Fridays closing up 1.5%+ are already followed by soft weeks (54.1% up, −0.11% mean), so most of the fade is a general give-some-back effect rather than something unique to Fed reversals.
The comparison that does survive is between the three ways a −1.5% Wednesday can resolve. All 148 of them since 1952:
| Resolution by Friday | n | +5d mean | +20d mean |
|---|---|---|---|
| Full recovery — this setup | 24 | −0.70% | −0.68% |
| Partial — Fri between Wed and Tue close | 52 | +0.20% | +1.60% |
| None — Fri at or below Wed close | 72 | +0.66% | +1.83% |
The V-shaped snapback is the worst of the three outcomes over the following month. Weeks that only partly recovered, and even weeks that stayed broken, went on to do better. The intuition isn't complicated: a two-day vertical repair spends the demand that a slower base-building process would have held in reserve.
One thing that does not happen is a full round trip back to the panic. Only 7 of the 24 traded back down to the Wednesday closing low within the next five sessions. Whatever the fade looks like, historically it has not been a retest.
About the “Fed Day” Part
This is the caveat that matters most, and the reason our screen is mechanical rather than Fed-specific. Of the 24 precedents, checked against the Federal Reserve's published meeting calendars, only two fell on an actual FOMC decision day.
One up, one down. That is not a sample; it is an anecdote with a standard deviation. Three more landed on the session immediately after a Tuesday FOMC decision — December 20, 2000, September 17, 2008 and August 10, 2011 — producing +0.71%, −3.82% and +2.18%. Add them and you get five events splitting 3–2. Still nothing.
Seven of the 24 predate 1994, when the Fed did not announce policy decisions on the day of the meeting at all, so a “Fed day” reaction is anachronistic for them. And the June 2026 occurrence in the table — June 10 — was not a Fed day either; that month's FOMC met on the 16th and 17th. What the data can speak to is the shape: a sharp midweek shock, fully repaired within two sessions. Whether it came from Powell, a CPI print, a bank failure or a war is not something 24 observations can separate.
Where We'd Push Back on Ourselves
Twenty-four observations is a small sample, and the sensitivity analysis is not reassuring. Move the Wednesday threshold to −2.0% and the Monday win rate drops to 50% (n = 10); move it to −1.25% and it drops to 51.4% (n = 37). A result that swings that much on a small change to an arbitrary parameter is a result to hold loosely. We picked −1.5% because it matches the event that prompted the question, not because it produced the best output — but the sensitivity deserves stating plainly.
We also examined multiple horizons and multiple cuts without correcting for multiple comparisons, so the single p-value below 0.05 — five-day underperformance against all sessions — should be read as suggestive rather than established, especially since it weakens to p = 0.149 against the more appropriate matched control.
Frequently Asked Questions
Does the market rise the Monday after a Fed-day plunge is erased by Friday?
It rose 14 of 24 times — 58.3%, mean +0.39%. But any Friday closing up 1.5%+ is followed by a 57.2% Monday, so the setup adds essentially nothing. Binomial p = 0.57 against that matched control.
How often does the S&P 500 erase a 1.5% Wednesday drop by Friday?
Rarely. Of the 148 qualifying −1.5% Wednesdays since 1952, only 24 — about one in six — had the whole loss erased by Friday's close. 52 partly recovered and 72 ended the week at or below the Wednesday close.
Has the pattern stopped working?
The record splits hard by era: 8-for-9 from 1990–2009, then 2-for-9 since 2010. Across the ten most recent occurrences Monday was up 3 times, median −0.29%. Nine observations per bucket is thin, but it cuts against the headline number.
Were these actually Fed days?
Mostly not. Only two of 24 fell on an FOMC decision day (November 12, 1997 and January 31, 2024), three more the session after a Tuesday decision, and seven predate 1994 when the Fed didn't announce on meeting day. A strict FOMC filter leaves a sample of two.
Is a Monday dip below Friday's close a warning sign?
No — it is the base case. 83% of these Mondays traded below Friday's close at some point, including most of the ones that closed green. Median worst drawdown −0.82%, median best extension +0.76%.
What happened on Monday, August 3, 2026?
It is not in the table above. Our price series for this study ends Friday, July 31, 2026 — the day the setup completed — so the most recent occurrence is scored through that close only. We'll add the outcome once the data is in.
What data did you use?
S&P 500 daily bars from February 21, 1952 to July 31, 2026 — 18,766 sessions after de-duplication. Returns are close-to-close. Sessions before 1962 carry no intraday range, so high/low statistics use the 1962+ subset; six Monday bars have an opening print equal to the prior close and are excluded from gap statistics.
This is the same failure mode we keep finding in market folklore: a real pattern, honest arithmetic, and a base rate nobody ran. See also the Mercury retrograde 2% drop claim, where the “danger dates” scored 17.8% against an 18.2% any-day base rate, and what happens to 100% win-rate seasonal patterns out of sample.
Patterns That Survive the Base-Rate Test
Seasonal Edge backtests every calendar and cycle pattern against decades of data — win rates, profit factors, and regime filters included — so you can see which recurring dates actually held up historically and which are folklore.
Free preview includes AAPL, NVDA, and TSLA. Cancel anytime.