UBER closing-half-hour (15:30–close) vs early-session (09:30–15:30): reversal or extension?
UBER’s close tends to nudge against the direction established earlier in the day, but the effect is subtle. Using minute bars over roughly three years (752 trading days), we asked whether the final 30 minutes (15:30–close) systematically reverses or extends the session move from 09:30–15:30.
The answer: mild reversal. The closing half-hour return is negatively correlated with the earlier move (r = −0.1299, p ≈ 0.0004) and the opposite-sign share is 53.19%. Mean closes are about −0.029% after earlier up-days and +0.0319% after earlier down-days, and the OLS slope implies roughly −2.9 basis points in the close for a +1% early-session move. It’s statistically real but explains only ~1.7% of variance (R² ≈ 0.0169). Full methodology, charts, and robustness checks follow below.
For UBER over the past ~3 years, does the closing 30 minutes (15:30–close) tend to reverse the direction of the rest of the session (9:30–15:30) — the late-day 'fade the day' rebalance — or pile on and extend the move into the bell? Thesis: the closing half-hour return is mildly negatively correlated with the earlier move, so strong up-days give a little back into the close and weak days bounce, a systematic reversal rather than momentum.
How this was measured
UBER minute bars filtered to US regular hours (Mon–Fri, 09:30–16:00 ET). For each trading day in the last ~3 years (anchored to the most recent session), computed the early-session move as close(15:30) / close(09:30) − 1 and the closing-half-hour move as last pre-16:00 close / close(15:30) − 1. Excluded early-close sessions (last timestamp before 15:55 ET) and days missing 09:30 or 15:30 bars. Measured Pearson correlation and an OLS slope of closing-half-hour vs early-session returns, and compared mean closing-half-hour returns conditioned on the early-session being up vs down, plus >±1% 'strong' days.
The key numbers
Reading the numbers
Across 752 trading days, the closing half-hour is mildly negatively correlated with the earlier session move (Pearson r = -0.1299, p = 0.0004), so there is a small but statistically detectable tendency for the close to fade the day's earlier direction.
The charts
The scatter of 752 sessions forms a cloud that slopes down slightly: early-session returns range from -0.1015 to 0.12 while closing 30m returns sit between -0.0174 and 0.0136. The OLS slope is -0.02877 and R² is about 0.0169, meaning the negative relationship is real but explains only ~1.7% of the variation. Look at the dense cluster around zero — most days show very small closing moves even when the early move is larger, so the fade exists but is modest and noisy.
The box comparison shows the average closing return is positive (mean = 0.0003) on the 396 days that were down earlier, and negative (mean = -0.0003) on the 356 earlier-up days. The ranges overlap substantially (earlier-down min -0.0147 to max 0.0136; earlier-up min -0.0174 to max 0.0126), so while the center shifts sign in the expected reversal direction, individual days often break the pattern. For the question at hand, the middle of the distributions supports a mild reversal but the spread warns against expecting it every day.
The bar chart quantifies the means: earlier down = 0.0003, earlier up = -0.0003, with stronger moves showing bigger means (strong down = 0.0006, strong up = -0.0005). The stronger-move groups (N=211 for strong down, N=195 for strong up) show about twice the average closing response compared with the full-direction groups, so large early moves tend to produce a slightly stronger fade into the close. Note that all these means are very small in absolute terms, so the effect is systematic but modest in magnitude.
Closing 30m return summary by early-session condition
| condition | n_days | mean_close30 | median_close30 | std_close30 | mean_early_move |
|---|---|---|---|---|---|
| Earlier down | 396 | 0.0003 | 0.0004 | 0.004 | -0.0136 |
| Earlier up | 356 | -0.0003 | -0.0001 | 0.0045 | 0.0147 |
| Strong down (<-1%) | 211 | 0.0006 | 0.0006 | 0.0041 | -0.0214 |
| Strong up (>+1%) | 195 | -0.0005 | -0.0003 | 0.0049 | 0.0228 |
The takeaway
Yes — the closing half-hour tends to mildly reverse the earlier session rather than extend it. The correlation is r = -0.1299 (p ≈ 0.00036) and the OLS slope is -0.02877, which translates to roughly -0.029% in the close for a +1% early-session move; 53.19% of days show opposite signs. In magnitude that means after early up-days the mean close is about -0.029% (N=356) and after early down-days it's about +0.032% (N=396); on strong days (>±1%) the means are roughly -0.0503% after strong up (N=195) and +0.0595% after strong down (N=211). The result is statistically robust (only about a 0.04% chance this pattern is pure luck) but tiny in economic terms — R² ≈ 0.0169, so the early move explains only ~1.7% of closing variation. Practical takeaway: there is a real lean toward late-day reversal, but it’s a small effect — on the order of 3–6 basis points on average, not a dependable standalone trade.
The fine print
- Sample excludes early-close days and sessions missing exact 09:30 or 15:30 bars; results apply to full-length sessions only.
- The late return uses the last pre-16:00 minute (auction prints at 16:00 are not explicitly captured).
- The effect is statistically significant but tiny (R² ≈ 0.0169) — high day-to-day noise swamps the signal.
- Findings are specific to UBER over the analyzed ~3-year window and may not hold for other tickers or regimes.