VIX closed at 14.53 on September 4, 2026 — near the bottom of its own 1-year range (13.47 low, 31.05 high), sitting at the 5.2th percentile of that range. Two stories get told about this every single time. One says it's complacency, a spike is overdue. The other says it's a healthy bull tape, ride it. Neither is usually checked against what actually happened the last 360 times VIX did this.
We ran the check. Since 1990, every time VIX entered the bottom decile of its trailing 1-year range for the first time (not counting repeat days inside the same stretch), SPX kept grinding higher more often than the unconditional baseline — but by less, not more. And VIX itself, the thing sitting at a low, was the more likely mover: it rose more often than it fell over the following 1-month to 12-month horizons, and rose more often than on a random day. The "calm before the storm" framing and the "ride the bull trend" framing are both half right, split by what you're actually watching.
What we measured
Signal: VIX's rolling 252-trading-day percentile rank drops to 10 or below (bottom decile of its own trailing 1-year range), on the first day of each new stretch — a day already inside the decile doesn't refire the signal, so 355-360 signals (count varies slightly by horizon due to data availability) are independent episodes, not one multi-week streak counted dozens of times. Entry is the next day's close, since the signal day's own close isn't tradable. Data: ^VIX and ^GSPC daily closes from yfinance, 1990-01-02 through 2026-09-04 — 9,237 trading days, index-level series with no reconstructed constituent list. 360 signals fire in that window: 84 in the 1990s, 100 in the 2000s, 117 in the 2010s, 59 in the 2020s so far — the pattern isn't a single-decade artifact.
Forward outcomes measured at 1-month (21 trading days), 3-month (63), 6-month (126), and 12-month (252) horizons, for both VIX itself and SPX, against the unconditional forward distribution over the same 1990-2026 window.
What the data says
SPX forward returns, signal vs baseline:
| Horizon | Signal mean | Baseline mean | Signal hit rate | Baseline hit rate |
|---|---|---|---|---|
| 1m | +0.46% | +0.82% | 64.2% | 64.0% |
| 3m | +1.92% | +2.44% | 73.5% | 70.2% |
| 6m | +4.63% | +4.89% | 79.7% | 74.5% |
| 12m | +8.64% | +10.20% | 83.8% | 80.7% |
The direction confirms the bullish story: hit rate is higher than baseline at every horizon, and climbs to 83.8% at 12 months. But the mean and median return are lower than baseline at every single horizon — 3-month median +2.79% vs baseline +3.17%, 12-month median +9.45% vs baseline +11.83%. A VIX bottom-decile day marks a market that keeps going up more reliably, but by less. That's the signature of a market that's already priced the calm in, not one about to accelerate.
VIX forward change, signal vs baseline:
| Horizon | Signal mean | Baseline mean | Signal hit rate (VIX up) | Baseline hit rate |
|---|---|---|---|---|
| 1m | +11.25% | +2.48% | 67.0% | 46.2% |
| 3m | +18.66% | +4.78% | 60.6% | 44.4% |
| 6m | +20.52% | +6.17% | 62.0% | 46.2% |
| 12m | +26.25% | +7.38% | 62.1% | 47.9% |
This is the sharper result. On a random day, VIX is a coin flip to be higher or lower a month later (46.2% baseline hit rate for "up"). From a bottom-decile day, it's higher 60.6-67.0% of the time, and the mean move sits well above the baseline mean at every horizon in the table above. VIX mean-reverts hard off its own lows — the "spike is coming" instinct is directionally correct far more often than a coin flip, even though the median move (+5.3% to +12.1%, depending on horizon) is a normalization, not a crash.
Threshold sensitivity (3-month SPX mean return): bottom 5th percentile +1.45%, 10th +1.92%, 15th +1.89%, 20th +1.71%, with signal counts of 295, 360, 375, and 400 respectively. The result holds across four different cutoffs instead of only existing at one hand-picked threshold.
The actionable rule
Split the question by what you're trading. If you're long equities, a VIX bottom-decile reading is not a sell signal — SPX hit rate is higher than baseline at every horizon out to a year, and the failure count below is small. But don't expect the low-VIX stretch to supercharge returns; size for baseline-or-slightly-below performance, not for acceleration. If you're trading volatility directly (VIX calls, VXX, a hedge book), the bottom decile is the better long entry: VIX rises more often than not and by a wide margin over the unconditional average at every horizon we tested (see the VIX table above), which is the opposite of what "VIX is cheap, stay short vol" usually implies.
Where it fails
23 of 355 checked episodes (6.5%) saw SPX draw down 10% or more within 3 months of the signal, worst case -31.48% (the January 9, 2020 signal, VIX at 12.54, run over by the COVID crash). Five of the eight worst failures cluster in two windows: December 2019-January 2020 (VIX 12.47-12.63 right before the crash) and April-May 2002 (VIX 18.11-19.35, dot-com bear market grinding on). One more is July 2011 (-18.2%, VIX at 15.95, debt-ceiling/US downgrade selloff). Regime split: 347 of 360 signals fired with SPX already above its 200-day moving average — this is overwhelmingly a bull-market phenomenon, and the 13 signals that fired below the 200dma are too few to draw a separate conclusion from. The lesson from the failures isn't that low VIX predicts the crash — the VIX level at each failure (12.54, 18.11-19.35) gave no advance warning of magnitude — it's that a low reading buys you better odds, not insurance.
How to watch it
The current reading — VIX 14.53, 5.2th percentile of its trailing 1-year range — sits inside the same bottom-decile zone this analysis is built on. Track the live level and regime classification on the Market Health dashboard.