Market risk · Senntric data study
Do Stock Market Crash Indicators Work? We Tested 14 Years of S&P 500 and Nasdaq Down Days
We tested the warning signs traders watch, from a low VIX to market breadth and options stress, against every big S&P 500 and Nasdaq-100 down day over 14 years. Most of the popular ones failed. A few shifted the odds, and none named the day.
Key findings
- 37% vs 20%SPY −2% day within 10 sessionsAfter a Stressed Market Health reading, against any day
- 55%Of early-fear alerts saw an SPY −2% dayWithin 10 sessions; 20% on any day. About 2.2 alerts a year
- 0.8×Edge from a low VIX in 2020–20261.0× is no better than a random calm day
- 63%Of Stressed days had no SPY −2% dayIn the next 10 sessions. Warnings are odds, not dates
01 / Short answer
Warning signs shift the odds. None name the day.
Across 14 years, the most useful warnings came from signs of stress that had already started to build, mostly in the options market. Signals of calm and complacency, the ones most often quoted before a crash, did not work.
Even the best warning was followed by a big down day only about half the time, and each one missed some of the sharpest drops. That makes these tools a guide to risk, not a sell signal.
02 / Market Health score
The score since 2012, with every big down day
The Market Health score combines five parts of the market into one 0–100 reading: risk appetite, breadth, credit, the S&P 500 trend and volatility. It spent 18% of days Stressed, 33% Weak, 31% Steady and 18% Strong. The red marks show every day SPY or QQQ closed down 2% or more.
Source: Senntric Market Health backtest on daily closes, May 2, 2012 to November 5, 2026.
Big down days cluster where the line sits low. They also appear from high readings: 18 SPY or QQQ −2% days came straight after a Strong score, including SPY’s -5.8% fall on June 11, 2020.
03 / Odds by zone
A Stressed score roughly doubled the odds of a big down day
After a Stressed reading, SPY had at least one −2% day within 10 sessions 37% of the time, against 20% on any day and 11% after a Strong reading. Larger drops followed the same pattern, and the Nasdaq-100 reacted more than the S&P 500.
SPY −2% day
Any day: 20%
SPY −3% day
Any day: 6%
QQQ −2% day
Any day: 34%
QQQ −3% day
Any day: 14%
Source: Senntric Market Health backtest on daily closes, May 2, 2012 to November 5, 2026.
| Event | Any day | Stressed | Weak | Steady | Strong |
|---|---|---|---|---|---|
| SPY −2% day within 5 sessions | 11.8% | 25.7% | 12.4% | 7.2% | 5.2% |
| SPY −2% day within 10 sessions | 19.5% | 37.0% | 21.0% | 13.2% | 11.1% |
| SPY −3% day within 5 sessions | 3.6% | 9.0% | 3.3% | 2.5% | 1.1% |
| SPY −3% day within 10 sessions | 6.1% | 13.3% | 5.8% | 4.1% | 3.5% |
| QQQ −2% day within 10 sessions | 34.0% | 57.5% | 38.2% | 25.3% | 19.2% |
| QQQ −3% day within 10 sessions | 14.3% | 33.0% | 12.4% | 11.2% | 5.3% |
The false alarms are the other half of the story. 63% of Stressed days had no SPY −2% day in the next 10 sessions, and 87% had no −3% day. The score turned Stressed 53 separate times, about 3.7 times a year, and SPY had a −2% day within 10 sessions after 14 of them.
04 / Calm markets
In a calm market, the score says little about the S&P 500
Part of the score’s edge comes from the market already wobbling: the trend and volatility parts fall as prices fall. Keeping only calm days, with no −2% day in the previous two weeks, removes that. On those days a Stressed reading barely moved SPY’s odds, 13% against 12%. QQQ’s still nearly doubled, 44% against 24%.
SPY −2% day, calm days
Any day: 12%
QQQ −2% day, calm days
Any day: 24%
Source: Senntric Market Health backtest on daily closes, May 2, 2012 to November 5, 2026.
So the score works best as a gauge of stress that has already set in. It rarely sounds the alarm while markets are quiet, and that is the question most traders have in mind when they ask about crash warning signs.
05 / What failed
Low VIX, put/call and breadth divergences did not hold up
We tested popular crash warning signs on calm days only, chose each setting on 2012–2019, then checked it unchanged on 2020–2026. An edge of 1.0× means alerts were followed by an SPY −2% day no more often than a random calm day. Most had no edge in one period or both, and the two that kept a number above 1.0× did so on very few alerts or faded on bigger selloffs.
| Warning sign | Edge 2012–19 | Edge 2020–26 | Alerts | Verdict |
|---|---|---|---|---|
| Very low VIX | 0.9× | 0.8× | 23 / 16 | No edge in either period |
| VIX bottoming and turning up | 0.0× | 0.0× | 12 / 9 | No alert was followed by a −2% day |
| Low put/call ratio | 0.7× | 1.6× | 16 / 19 | Unstable: flipped between periods |
| Breadth divergence at highs | 3.5× | 2.1× | 6 / 3 | Too few alerts later; 1 of 22 versions held up |
| 40-input statistical model | 2.9× | 1.0× | 29 / 25 | Overfit: a coin flip on new years |
| Negative dealer gamma | 0.0× | 0.9× | 3 / 7 | No calm-day edge, few alerts |
| Weak dark-pool buying | 1.8× | 2.8× | 18 / 11 | Modest; faded on bigger selloffs |
Source: Senntric Market Health backtest on daily closes, May 2, 2012 to November 5, 2026.
- Complacency did not predict crashes. A very low VIX and a VIX turning up from a low had no edge in either period. A low put/call ratio flipped from worse than random to better, which is not a signal you can lean on.
- Breadth divergences failed out of sample. The best version on 2012–2019 fired only 3 times in 2020–2026. Only one of the 22 versions we tried kept an edge above 1.3× in both periods.
- More inputs made it worse. A statistical model on 40 inputs looked strong on the years it learned from (2.9×) and was a coin flip on new years (1.0×).
- Dealer gamma and dark-pool buying were weak. Negative dealer gamma had no calm-day edge. Weak dark-pool buying showed some lift in both periods but faded on bigger selloffs, so we count it as modest, not reliable.
06 / Early fear
The best warning came from options-market stress
The strongest signal we found reads stress in options prices before it shows up in stocks. We call it early fear. It needs four signals to turn on together:
- Near-term fear: the 9-day VIX above the 30-day VIX.
- Volatility of the VIX: a high VVIX.
- A jump in implied correlation: traders pricing stocks to fall together over the next month.
- Inverted correlation: 1-month implied correlation above 3-month.
Early fear fired 31 times from 2012 to October 2026, about 2.2 times a year. SPY had a −2% day within 10 sessions after 55% of alerts, against 20% on any day, and a −3% day within 5 sessions after 23%, against 4%.
SPY −2% day within 10 sessions
Any day: 20%
SPY −3% day within 5 sessions
Any day: 4%
Source: Senntric Market Health backtest on daily closes, May 2, 2012 to November 5, 2026.
The combination is what matters. Two or three signals on together fired 5.3 times a year with almost no edge (23% against 20%). On their own, the four signals had edges of 1.3× to 1.9× in 2020–2026.
Source: Senntric Market Health backtest on daily closes, May 2, 2012 to November 5, 2026.
It missed some of the sharpest drops. It fired only on or after the first big down day in the August 2024 yen unwind, the December 2024 Fed day and the April 2025 tariff selloff, and it was silent before the August 2022 and June 5, 2026 drops.
07 / November 2026
How the November 2026 selloff looked from the inside
The score was 63 at the October 6 high and fell to 29 by October 14. It stayed Stressed on 11 of the 16 sessions to November 4. QQQ slid early, while SPY’s first −2% day did not come until November 5, when SPY fell 4.4% and QQQ 6.5%. Early fear fired on November 3, two sessions before.
Source: Senntric Market Health backtest on daily closes, May 2, 2012 to November 5, 2026.
This is one selloff, and the early-fear alert came after the test window, so it is an example rather than evidence. It does show the pattern from the backtest: the score gauged rising stress for weeks, and the options-market signal fired close to the drop.
08 / Using it
How to use crash warning signs
- Read warnings as odds. A Stressed market means big down days are about twice as likely as usual. It does not mean one is coming this week.
- Size for the zone. When stress builds, smaller positions and wider stops fit the higher chance of a large move. That works whether or not the drop comes.
- Watch options stress, not complacency. A calm, low VIX told us nothing. Short-dated fear rising above longer-dated fear, together with correlation jumps, told us more.
- Check the calendar for event risk. Drops set off by a central-bank decision or a policy announcement come without warning in these signals. Know when those events are scheduled, and size for them separately.
- Expect misses. Big down days still came from Strong readings, and the best signal missed several selloffs. Keep a plan for the drop no indicator flags.
On Senntric you can watch the volatility indexes on the chart next to SPY gamma levels and the day’s market news, so stress is visible as it builds.
09 / In Senntric
Market Health and early fear, live in Senntric
The score and the early-fear signals from this study run live in Senntric’s Market Health panel. It shows today’s score and zone, the five parts behind it, the odds of a big down day from the current zone, and which of the four options-stress signals are on. Pro members can open it from the command bar by typing “market health”.
Stressed
▲ 1 from the previous session (19)
Odds of an S&P −2% day within 2 weeks
Big down days are about twice as likely as usual. Risk is high, but it doesn’t time the drop. Line: 20% on any day.
What stands out
- The ratio of high-beta to low-volatility stocks is 15.5% below its 50-day average, so buyers favour defensive stocks.
- The volatility part dropped 17 points in five sessions, the biggest move of the five.
- Breadth is the strongest part at 93, which is holding the score up.
Components
5-day change
Score history · 6 months
Senntric Market Health panel, session of November 6, 2026. Recreated from the live panel.
The panel tracks every new early-fear alert and marks whether a big down day followed, so the live record builds alongside the backtest.
FAQ
Questions about stock market crash indicators
Do stock market crash indicators work?
Some do, weakly. In our 2012–2026 backtest, a low Market Health score roughly doubled the odds of an S&P 500 (SPY) drop of 2% or more within two weeks (37% against 20% on any day), but about two in three warnings passed with no such day. Popular complacency signals such as a low VIX had no edge at all.
Does a low VIX predict a stock market crash?
Not in our test. Alerts when the VIX sat near the bottom of its one-year range were followed by an SPY −2% day no more often than a random calm day (0.9× in 2012–2019 and 0.8× in 2020–2026, where 1.0× means no edge). The VIX turning up from a low did no better.
What is the best leading indicator of a stock market drop?
The strongest signal we found was options-market stress: near-term VIX above 30-day VIX, a high VVIX, and a jump in short-dated implied correlation, all at once. It fired about 2.2 times a year and was followed by an SPY −2% day within 10 sessions 55% of the time, against 20% on any day. It was chosen with recent years in view, so live tracking is the real test.
Can you predict a stock market crash?
Not reliably. The best signals in this study shifted the odds of a big down day, but none named the day, and each one missed some of the sharpest selloffs, including the August 2024 yen unwind and the April 2025 tariff drop. Those were event-driven: they began with sudden news, a central-bank move or a tariff announcement, that left no build-up of stress to detect beforehand. Treat warning signs as a reason to manage risk, not as a timing tool.
Method
How we measured it
Period and markets. 3,650 sessions from May 2, 2012 to November 5, 2026. Down days are daily close-to-close moves in SPY (S&P 500) and QQQ (Nasdaq-100). The start date is the earliest the score’s inputs allow.
The Market Health score. The score Senntric runs today, with its rules unchanged: five parts (risk appetite measured by high-beta against low-volatility stocks, market breadth, credit, the S&P 500 trend, and volatility) combined into one 0–100 reading. Zones: Stressed under 30, Weak 30–49, Steady 50–69, Strong 70 and above.
Odds. For each zone, the share of days followed by at least one drop of the stated size within the next 5 or 10 sessions, against the same share for all days. “Calm days” keeps only days with no SPY or QQQ −2% day in the previous 10 sessions, which removes readings taken after a selloff has already begun.
Signal search. Each idea’s setting was chosen on 2012–2019 and then checked unchanged on 2020–2026. Edge is the share of alerts followed by an SPY −2% day within 10 sessions, divided by the same share on calm days in that period. Alerts count once per episode. We publish the signal types but not their exact settings.
Early fear. An alert needs all four options-stress signals on together. Its outcomes use alerts through October 8, 2026, so that every alert has a full window after it. The November 3, 2026 alert is shown but not counted.
Sources
Sources and disclaimer
- Senntric Market Health score
The production score rules, run on stored daily closes for each input. - Senntric stored daily closes
SPY, QQQ, the index and ETF inputs, and the volatility and implied-correlation indexes (VIX, VIX9D, VVIX and implied correlation). - Study snapshot
The rows behind every figure on this page, frozen on November 7, 2026.
Fourteen years hold only 109 SPY −2% days and a handful of true crashes, so small changes in the rules move the percentages. The early-fear set was chosen with 2020–2026 in view, which flatters its record; we started tracking it live on November 6, 2026. This research is educational. It is not a forecast, a recommendation, or investment advice, and past moves do not predict future ones.
Senntric
See the stress build while it is happening
Senntric is a trading terminal for active traders: charts, live market news and audio squawk, options positioning, Market Health with early fear, and an AI analyst that reads the market with you. Join to follow the next selloff as it develops, not after it.