HOOD: Next-day return after up-days — does high volume ‘confirm the move’? (last ~3 years)
High-volume up-days don’t meaningfully “confirm” HOOD’s move. We examined roughly 36 months of minute bars aggregated to daily OHLCV and isolated 372 up-days, splitting extremes into 124 high-volume and 124 low-volume up-days. The average next-day return gap between high and low volume was only about 0.12 percentage points — far too small to signal a reliable edge.
Across means, medians, and nonparametric tests the differences look like sampling noise rather than an exploitable pattern. The full dataset, distributional charts, and hypothesis tests are below; the short takeaway is simple: for HOOD over this period, volume behind an up-day does not provide a consistent next-day directional edge.
For HOOD over the past ~3 years, do up-days on heavy volume actually keep climbing the next session while up-days on light volume fade — the classic 'volume confirms the move' belief — or does the volume behind today's move carry no edge for next-day direction? Thesis: next-day returns are essentially independent of whether today's move came on high or low volume, so 'volume confirmation' is a story technicians tell after the fact rather than a tradeable signal.
How this was measured
Resampled HOOD minute bars to daily OHLCV over the last ~36 months. Computed close-to-close daily return and next-day return. Defined relative volume (rvol) as today's total volume divided by the trailing 20-trading-day average volume ending the PRIOR day (point-in-time). Restricted to UP days (ret>0), then split those days into Low/Mid/High rvol terciles (in-sample) and compared next-day return distributions. Reported mean gap (High−Low), fraction-positive, Welch t-test (means), and Mann–Whitney U (medians).
The key numbers
Reading the numbers
Across 372 up-days, the high-volume and low-volume buckets each contain 124 days. Mean next-day returns are -0.00026806044231204726 (High) vs -0.0015045832129548073 (Low), gap 0.00123652277064276 and p-value 0.8038537908280513 → no clear mean difference.
The charts
The box plots show the full spread of next-day returns for low, mid and high volume up-days. The mid-volume bucket actually has the only positive mean (0.0048) while low and high means sit near zero (−0.0015 and −0.0003), and all three buckets have wide overlapping tails (mins down to about −0.13 and highs up to about 0.17). Look at the overlap in the boxes and tails rather than the small differences in means — volume tercile doesn't produce a clean, monotonic shift in next-day returns.
The bar chart highlights mean next-day returns by bucket: Low = −0.0015, Mid = 0.0048, High = −0.0003. The mid bucket is the only one slightly positive, but the differences are tiny and the statistical test reported (two-sided Welch p = 0.8039) says we cannot treat these mean gaps as reliable. In plain terms, the average next-day move after a high-volume up-day is essentially indistinguishable from a low-volume up-day.
The scatter of rvol versus next-day return is a cloud with mean rvol 1.0647 (range 0.3075–7.6101) and next-day returns centered near 0.001, showing no visible slope. The Pearson r is 0.028334124120412037 (|r| ≈ 0.028), which is vanishingly small and indicates a negligible association. In other words, higher relative volume on the up-day does not produce a consistent next-day gain across these 372 observations.
Next-day return summary by UP-day volume bucket
| bucket | n | mean | median | std | frac_positive |
|---|---|---|---|---|---|
| Low volume | 124 | -0.0015 | -0.0039 | 0.0355 | 0.4516 |
| Mid volume | 124 | 0.0048 | 0.005 | 0.0465 | 0.5403 |
| High volume | 124 | -0.0003 | -0.0014 | 0.0425 | 0.4839 |
The takeaway
No — for HOOD over the last ~3 years, the volume behind an up-day does not provide a reliable next-day edge. Across 372 up-days (124 high-volume and 124 low-volume), the mean next-day return was about -0.03% after high-volume up-days versus about -0.15% after low-volume up-days, a gap of roughly +0.12 percentage points. In raw success rates the difference is small: 48.4% of high-volume up-days closed up the next day versus 45.2% for low-volume up-days. Those tiny differences are statistically indistinguishable: Welch two-sided p ≈ 0.804 (about an 80-in-100 chance this difference could be sampling noise), Mann–Whitney p ≈ 0.631, and the Pearson correlation r ≈ 0.028 — essentially no relationship. In short, this looks like a coin flip, not a tradeable signal. Practical takeaway: don’t count on “volume confirms the move” to predict HOOD’s next-day direction over this period.
The fine print
- Volume buckets are in-sample terciles (descriptive); live thresholds would drift and need out-of-sample testing.
- Relative volume uses a 20-day trailing average anchored to the prior day; the first ~21 trading days were dropped.
- Next-day returns are close-to-close and include overnight gaps; intraday continuation or same-day signals aren’t tested here.
- Financial returns show autocorrelation/volatility clustering, so standard p-values can be optimistic.