On Thursday, November 5, SPY fell 4.41% to 733 and VIX went from 16.61 to 28.18 in one session, a +69.66% jump. Since SPY started trading in 1993, only 3 days have produced a VIX jump that large in percentage terms, and the SPY drop ranks 36th worst out of 8,501 sessions. The two takes that show up the morning after are always the same: "big down days get bought" or "this is the first crack, more to come".
We checked both against every comparable session since 1993. The short answer: the next-day bounce is a slight edge on paper (59.22% of 3%+ drop days closed higher the next day, vs 54.02% on any random day) but it has faded to a coin flip in the 2010s and 2020s. The one thing the data says with confidence is that tomorrow will be big. After a 3%+ drop, SPY's average next-day move, up or down, is 2.25%, against 0.78% on a normal day.
What we measured
Data: SPY daily OHLC (split- and dividend-adjusted) and ^VIX daily closes from yfinance, 1993-01-29 through 2026-11-05, 8,501 trading days. ETF and index series only, so no survivorship problem.
Four signal definitions, all known at the signal-day close:
- Big drop: SPY close-to-close return of -3% or worse.
- VIX spike: VIX close-to-close change of +40% or more.
- Drop and spike: both a 3%+ SPY drop and a VIX jump of +30% or more.
- Drop from calm: a 3%+ SPY drop when VIX closed under 20 the day before. This is the closest analog to November 5, when VIX had been at 16.61.
"Next day" is the signal-day close to the next close. We also split it into the overnight gap (close to next open) and the intraday session (next open to next close), plus 5-day and 21-day forward returns. Signals within 5 trading days of each other are counted as one episode, and we report both counts. The baseline is every trading day in the same window.
What the data says
Next-day SPY return after each signal, vs baseline:
| Signal | Days | Mean | Median | Closed up |
|---|---|---|---|---|
| Any day (baseline) | 8,500 | +0.05% | +0.07% | 54.02% |
| SPY -3% or worse | 103 | +0.50% | +0.38% | 59.22% |
| VIX +40% or more | 21 | +0.73% | +0.47% | 61.9% |
| SPY -3% and VIX +30% | 23 | +0.60% | -0.15% | 47.83% |
| SPY -3% from VIX under 20 | 11 | -0.39% | +0.24% | 54.55% |
Three things stand out.
The bounce is real but thin, and the mean flatters it. A 3%+ drop day is followed by a higher close 59.22% of the time, 5.2 points better than a random day. The mean of +0.50% is ten times the baseline, but the median is +0.38%, and 42 of the 103 next days were down anyway. Counting each selloff cluster once (62 episodes) the first-day hit rate is 64.52%, so the edge is slightly better on the first shock than on the follow-on days of a crash.
When the drop comes with a VIX spike, the bounce edge disappears. Days where SPY fell 3%+ and VIX jumped 30%+ closed higher the next day only 47.83% of the time, below the baseline, with a negative median of -0.15%. That is the bucket November 5 falls in. A big VIX jump means the options market is repricing risk fast, and that repricing often isn't done in one session.
Out of calm, the gap is the danger. In the 11 cases where SPY dropped 3%+ from a VIX under 20, the next morning gapped down 54.55% of the time (mean gap -0.66%), but the intraday session recovered: open-to-close was up 63.64% of the time, mean +0.28%. If you sell, the open has historically been the worst price of the day more often than not.
The size of tomorrow is the one reliable number. After a 3%+ drop, 38.83% of next days moved 2% or more in either direction. On a random day that happens 7.2% of the time. The 5-day path is rough too: the average worst low over the next 5 sessions is -4.68% below the signal close, vs -1.63% for a random day.
A week out, the odds lean up. 5-day returns after a 3%+ drop averaged +0.93% (up 62.14% of the time) against +0.23% (up 58.65%) at baseline. The drop-from-calm cases were up 5 days later 72.73% of the time, mean +0.61%. VIX spikes of +40% were followed by a higher SPY 21 days later 80.95% of the time vs 65.71% at baseline.
If it gaps up the next morning, should you buy it?
This is the trade most people actually face: the futures are green, SPY opens above yesterday's close, and it feels like the low is in. We tested buying the next open after every 3%+ drop and holding to the close 1, 2, 3, 5, 10 and 21 sessions later (hold 1 means exit at the same day's close), split by whether that open gapped up or down. 56.31% of the 103 drop days were followed by a gap-up open.
Buy the next open after a 3%+ SPY drop, mean return and % positive:
| Hold | Any open (baseline) | Gap-up open (58) | Gap-down open (45) |
|---|---|---|---|
| Same-day close | +0.01%, 52.34% | +0.52%, 65.52% | +0.54%, 60.0% |
| 2 days | +0.05%, 54.87% | -0.17%, 53.45% | +1.49%, 62.22% |
| 3 days | +0.1%, 56.53% | +0.03%, 55.17% | +1.27%, 57.78% |
| 5 days | +0.19%, 57.87% | -0.02%, 46.55% | +2.27%, 71.11% |
| 10 days | +0.42%, 61.22% | -0.2%, 55.17% | +1.65%, 68.89% |
| 21 days | +0.92%, 65.45% | +0.56%, 60.34% | +3.7%, 66.67% |
The gap-up open is a day trade, not a swing entry. Bought at the open and sold at that day's close, it made money 65.52% of the time. Held any longer, it did worse than buying a random open: flat to negative through 10 days, and positive only 46.55% of the time at 5 days. The gap-down open is the better entry at every hold from 2 days out, +2.27% at 5 days with 71.11% positive.
In the drop-plus-VIX-spike bucket, which is where November 5 sits, the split is sharper. Gap-up opens (12 cases) averaged -1.89% after 2 days and -2.37% after 5, positive only 33.33% of the time at both. The 11 gap-down opens in the same bucket averaged +4.38% at 5 days and every one of them was positive. Twelve and eleven cases are small, so read this as direction, not precision, but it points the same way as the 103-day set.
The likely reason: a gap-up after a VIX shock is short covering and dealer hedges unwinding, and it uses up the buying that would otherwise carry a bounce. A gap-down open is where the forced selling finishes.
The bounce has been fading
Split the 3%+ drop days by decade and the next-day bounce looks very different:
| Decade | Drop days | Next-day mean | Closed up |
|---|---|---|---|
| 1990s | 9 | +1.63% | 77.78% |
| 2000s | 48 | +0.63% | 64.58% |
| 2010s | 20 | +0.2% | 50.0% |
| 2020s | 26 | +0.1% | 50.0% |
Most of the headline edge comes from the 1990s and the 2008 crash cluster. Since 2010, the next day after a 3%+ drop has been a coin flip with a tiny positive mean. "Buy the next morning" is not a rule the recent data supports.
The threshold choice doesn't change the story much. At -2.5% (183 days) the next-day up rate is 59.02%; at -3.5% (65 days) it is 60.0%; at -4% (48 days) 58.33%. Deeper drops have slightly bigger mean bounces (+0.84% at -4% vs +0.36% at -2.5%), but the hit rate barely moves.
The 11 closest analogs
Every 3%+ SPY drop that started with VIX under 20 the day before:
| Date | SPY | VIX change | Next day | Next 5 days | Next 21 days |
|---|---|---|---|---|---|
| 1996-03-08 | -3.24% | +25.53% | +1.16% | +1.43% | +1.46% |
| 1997-04-11 | -3.18% | +3.95% | +1.34% | +4.34% | +14.48% |
| 2007-02-27 | -3.91% | +64.22% | +1.03% | +0.14% | +2.07% |
| 2008-06-06 | -3.19% | +26.46% | +0.24% | -0.1% | -6.17% |
| 2015-08-21 | -3.01% | +46.45% | -4.21% | +0.73% | -1.47% |
| 2016-06-24 | -3.59% | +49.33% | -1.79% | +3.29% | +6.65% |
| 2018-02-05 | -4.18% | +115.6% | +1.97% | +0.53% | +3.35% |
| 2018-10-10 | -3.17% | +43.95% | -2.2% | +0.77% | +0.79% |
| 2018-12-04 | -3.24% | +26.16% | -0.15% | -1.77% | -5.32% |
| 2019-08-05 | -3.01% | +39.64% | +1.4% | +1.5% | +3.6% |
| 2020-02-24 | -3.32% | +46.55% | -3.03% | -4.13% | -24.14% |
The closest single match on VIX behaviour is February 5, 2018 (Volmageddon): VIX more than doubled from 17.31, SPY fell 4.18%, and the next day closed +1.97% after opening down 1.51%. The closest match on drop size from a calm start is February 27, 2007, which bounced +1.03% the next day and was roughly flat a week later.
The rule
Don't treat the morning after a 3%+ drop with a VIX spike as a buy signal. Since 2010 the next-day odds are 50/50, and in the drop-plus-spike bucket they are slightly worse than a normal day. What the data does support:
- Size for a 2%+ day either way. Cut position size so a move of 2.25%, the average next-day absolute move, doesn't force a decision.
- Don't sell the open in a panic. Out-of-calm selloffs gapped lower 54.55% of the time but recovered intraday 63.64% of the time.
- Don't chase a gap-up open for a swing. If SPY gaps up the next morning, it is a same-day trade at best. Held 5 days, the gap-up entry after a 3%+ drop has been positive 46.55% of the time, worse than any random open. If you want a multi-day entry, a gap-down open has been the better one.
- Judge the week, not the day. 5 days later, SPY was higher in 8 of the 11 calm-start cases.
Where it fails
The bounce failed 42 times out of 103 after 3%+ drops, and 10 of those next days fell another 3% or more. Among the 11 calm-start analogs, 5 closed lower the next day, the worst being August 24, 2015, which gapped down 5.23% at the open and closed -4.21%.
The worst case is the one everyone is thinking about: February 24, 2020. SPY fell 3.32% out of a VIX of 17.08, dropped another 3.03% the next day, hit a low 11.44% below the signal close within 5 sessions, and was down 24.14% a month later. Nothing in the signal-day data distinguished it from 2018 or 2019. That is the point: one-day price and VIX moves tell you volatility is coming, not which direction it resolves.
Small-sample warning: the drop-from-calm bucket is 11 days and the drop-and-spike bucket is 23. Treat those numbers as suggestive. The 103-day big-drop set is the one with real weight.
How to watch it
The question that matters after day one is whether breadth confirms the selloff or diverges from it. Track that on Market Health, where the regime and breadth readings update every session.