QUANT RESEARCH · 量化研究
You don’t need the perfect moment.
VIX 30 / SPY × regular investing. No perfect entry is required, and the VIX 30 signal showed higher average subsequent returns in our sample.
MY TAKEAWAY
Two conclusions from this research.
1. We don’t need a perfect moment to invest. Regular investing finished ahead under these accumulation rules, even though the fear signal identified opportunities with higher average subsequent returns. Waiting for the signal need not be a prerequisite for putting new money to work.
2. VIX 30 worked as an entry signal in this historical sample. Here, “worked” means SPY had a higher average subsequent return after the signal than after the matched moderate-fear events. A useful entry signal and not needing a perfect entry can both be true. This does not establish guaranteed rebounds, future outperformance, or a complete entry-and-exit strategy.
QUESTION 01 · AFTER A FEAR SIGNAL
SPY performed better on average after VIX 30.
The signal is a VIX close reaching or exceeding 30 after a previous observation below 30—not a reading of exactly 30. The event study retained 48 crossings after a 21-session cooldown, with entry at the next SPY open. The 21-session average price return was +2.41%, versus +0.84% for matched moderate-fear events: a +1.57 percentage-point historical advantage.
THE RESULT, WITH ITS UNCERTAINTY
Give the signal a little time.
95% bootstrap interval for the mean: +0.65% to +4.08%
Historical event study, 1993–2026. Price returns; dividends not reinvested. This interval describes uncertainty in the historical mean—not the range of a future trade. Exploratory testing and crisis dependence remain limitations.
Mean 21-session adverse excursion: −3.56%. Worst close-based adverse excursion: −24.49%. A positive average did not make every entry safe. The matched-return difference’s 95% interval was −0.44 to +3.53 percentage points and includes zero: the higher sample average does not establish statistically conclusive outperformance against that benchmark.
QUESTION 02 · THE WHOLE JOURNEY
Count the cost of waiting.
Same initial investment. Same monthly additions. A buys every month; B waits in cash until the previous trading day’s VIX close is at least 30. Both hold SPY after buying.
| Metric | Monthly investing | VIX ≥ 30 |
|---|---|---|
| Cumulative time-weighted return | 1634.41% | 1471.07% |
| Annualized time-weighted return | 8.88% | 8.55% |
| Maximum drawdown (TWR curve) | −56.47% | −56.36% |
| Average cash allocation | 0.00% | 3.78% |
| New-cash deployment days (initial purchase excluded) | 403 | 78 |
| Mean wait for deployed contributions | 0 days | 405 days |
In this comparison, maximum drawdown is −56.47% for regular investing and −56.36% for VIX 30. One possible explanation is that we set no exit point. Both begin fully invested and only the timing of new contributions differs; purchased SPY remains invested through subsequent market declines.
This is a possible explanation, not a tested causal conclusion. We did not compare strategies with and without exits. Similar drawdowns neither invalidate the entry signal nor establish that adding take-profit or stop-loss rules would reduce drawdowns. The event study’s 1-, 5- and 21-session measurement windows are not sell rules in this accumulation test.
Waiting time covers the 398 monthly contributions already deployed; 5 were still waiting at the cutoff and are excluded from that completed-wait average. The longest completed wait was 1,730 calendar days. These 78 deployment days use different rules from the 48 event-study signals.
One threshold isn’t the whole answer.
The original workbooks tested 25 and 35. I reproduced both and added 30 using the same underlying observations and contribution schedule. The result is not monotonic: a higher threshold did not simply make outcomes worse.
| Strategy | Annualized TWR | Ending value vs. DCA | Cash deployment days |
|---|---|---|---|
| VIX ≥ 25 | 8.45% | −7.30% | 139 |
| VIX ≥ 30 | 8.55% | −5.63% | 78 |
| VIX ≥ 35 | 8.63% | −3.72% | 40 |
Sensitivity table: cash earns the prior available 3-month Treasury yield proxy. All three use identical monthly additions. Initial purchase excluded from deployment counts.
The rules behind the numbers.
A fair cash-flow comparison
Both start fully invested. From the next month, a fixed addition equal to 1% of the original principal arrives on the first SPY trading day of each month. There are 403 additions. Only the new contributions are timed; this is not a test of delaying the initial lump sum.
No future information
VIX and cash yields use prior available observations. Cash accrues over calendar days before that day’s contribution. Purchases occur at the open; valuation is at the close. Time-weighted returns split each contribution day at the open, then chain the subperiod returns.
Different questions, different rules
The event study tests a crossing, a cooldown and a fixed holding window. The accumulation test checks VIX ≥ 30 whenever cash is available and never sells. Do not compound the event average or interpret it as the long-term strategy return.
What this cannot settle
Historical price-only data; no dividends, fees, taxes, slippage or fractional-share constraints. Cash yield is an idealized proxy. Rules were explored on known data; there is no untouched out-of-sample validation. Signal frequency, entry timing and cash drag all matter. The result does not isolate frequency as the sole cause.