SPY forward 5d/10d returns after top-decile vs bottom-decile intraday range (≈3y window)
Do the spikiest SPY sessions actually signal short-term buying opportunities? We tested that exact hypothesis over roughly three years by using each day’s intraday high−low range divided by close as a VIX-like proxy, then comparing forward 5- and 10‑day close-to-close returns after the highest-range and lowest-range decile days.
The result is suggestive but weak. Top-decile range days averaged +0.71% over the next 5 trading days (vs an unconditional +0.43%), while bottom-decile (calm) days averaged only +0.21% — a “fear pays / calm lags” pattern. However, tail sample sizes are small (~75 days), dispersion is large, and p-values (top-decile vs rest: 0.293 for 5d, 0.601 for 10d) leave the edge far from conclusive. Full methodology, stats, and charts follow below.
For SPY over the past ~3 years, do the most volatile sessions — measured by SPY's own intraday high-low range as a VIX proxy — actually mark buying opportunities, with forward 5- and 10-day returns running higher after high-range 'fear' days than after the calmest ones? Thesis: forward returns following the top-decile volatility days meaningfully beat the baseline while ultra-calm 'complacency' days deliver flat-to-soft returns, showing that fear pays and calm doesn't.
How this was measured
Resampled SPY minute bars to daily OHLCV, computed an intraday realized-range proxy as (high−low)/close. Classified days into top decile (highest range), bottom decile (lowest range), and middle 80% using in-sample deciles over the ~3-year window ending at the latest available date. Computed forward 5- and 10-trading-day close-to-close returns from the day of the signal (close[t+H]/close[t]−1). Compared mean forward returns across buckets and ran Welch two-sample t-tests: top-decile vs rest and bottom-decile vs rest for each horizon.
The key numbers
Reading the numbers
Top-volatility days average +0.71% over 5 trading days and +1.00% over 10 days, versus bottom-decile +0.21% (5d) and +0.50% (10d). The top-decile edge versus the unconditional mean is about +0.28% (5d) and +0.16% (10d).
The charts
This bar chart shows mean forward returns for three realized-range buckets: calm (bottom decile), middle 80%, and high-volatility (top decile). The right-hand bars (top decile) are the tallest for both horizons — 0.71% over 5 days and 1.00% over 10 days versus 0.21% and 0.50% in the bottom decile — so on average the highest-range days lead to bigger short-term gains. That gap is the simple 'fear pays' signal: high-range days beat the unconditional mean by about 0.28% (5d) and 0.16% (10d), though these are average effects, not guarantees for each instance.
This box plot for forward 5-day returns gives the distribution behind the means. The top-decile group (n=75) has the highest mean (0.71%) but also much wider tails (min −8.51% and max +8.77%) than the calm group (n=75, mean 0.21%, min −3.41%, max +2.97%). What to watch: the top-decile’s larger upside is paired with larger downside risk, so while fear days raise the average forward return they also produce more variable outcomes than calm days.
The forward 10-day box plot tells the same story on a slightly longer horizon. The top-decile mean is 1.00% (n=74) compared with 0.50% for the bottom decile (n=75) and 0.86% for the middle 80% (n=593). Range-wise the top decile’s outcomes sit between the middle group’s widest negatives and the bottom group’s more constrained moves (top decile min −4.27% max +9.39%), so the 10-day lens still shows higher average returns after high-range days but with overlapping distributions and meaningful tail risk.
Per-bucket forward-return summary
| bucket | N_5d | mean_5d | median_5d | std_5d | frac_pos_5d | N_10d | mean_10d | median_10d | std_10d | frac_pos_10d |
|---|---|---|---|---|---|---|---|---|---|---|
| Bottom decile (calm) | 75 | 0.0021 | 0.0035 | 0.0116 | 0.5867 | 75 | 0.005 | 0.0066 | 0.0177 | 0.6133 |
| Middle 80% | 597 | 0.0042 | 0.0055 | 0.0197 | 0.6248 | 593 | 0.0086 | 0.0117 | 0.0278 | 0.6813 |
| Top decile (high vol) | 75 | 0.0071 | 0.008 | 0.0251 | 0.6933 | 74 | 0.01 | 0.007 | 0.0281 | 0.6081 |
The takeaway
Yes — the data show a small average edge after the highest-range SPY days and weaker returns after the calmest days, but the evidence is thin. Top-decile days averaged +0.71% over the next 5 trading days (N=75) and +1.00% over 10 days (N=74), versus unconditional means of +0.43% (5d) and +0.84% (10d) — edges of about +0.28 and +0.16 percentage points. Bottom-decile (calm) days averaged +0.21% (5d) and +0.50% (10d), underperforming the unconditional by roughly -0.22 and -0.33 percentage points. Those gaps are not statistically robust: top-decile vs rest has p=0.293 (5d) and p=0.601 (10d) — roughly a 29-in-100 and 60-in-100 chance those top-decile edges are just noise; bottom-decile p≈0.11 (~11-in-100 chance) is a weak lean but still not definitive. Dispersion is large relative to the edges (top 5d std ≈ 2.51%, bottom 5d std ≈ 1.16%), and sample sizes in each tail are only ~75 days, so results are noisy. Practical takeaway: there is a suggestive “fear pays / calm lags” pattern on average, but it’s far from conclusive — treat it as a tentative signal that needs stronger, controlled testing before trading on it.
The fine print
- Decile cutoffs are computed in-sample over the ~3-year window; out-of-sample behavior may differ.
- Forward 5d/10d returns overlap across signal days, which inflates uncertainty and can make p-values optimistic.
- (high−low)/close is an intraday-range proxy, not VIX; results are proxy-dependent.
- No controls were applied for macro regime, Fed/earnings days, or other calendar clustering that could drive the effect.