AI Research UBER

UBER 'power hour' vs midday: share of daily range and contribution to the day's return (≈3y)

Surprising but clear: UBER’s final hour is louder on a per‑hour basis, yet it doesn’t reliably decide the day’s direction. Over the past ~three years we compared the 15:00–16:00 window to the 12:00–14:00 midday stretch (normalized per hour) and found the close accounts for about 28.98% of the day’s high–low range versus 20.67% for a midday hour.

The study uses minute bars through regular US hours and a paired t‑test to compare per‑hour range shares; charts and correlation tests then examine directional alignment. The range-concentration result is robust, but the lunchtime moves actually correlate more with the day’s net open‑to‑close returns. Read on for the full statistics, charts, and sign‑alignment analysis.

The research question

For UBER over the past ~3 years, does the 'power hour' actually run the tape — does the final 60 minutes (15:00–16:00 ET) generate a disproportionate share of the day's total high-low range and of its net open-to-close return versus the sleepy midday stretch (12:00–14:00)? Thesis: the closing hour concentrates the day's volatility and decides its direction as closing-auction and rebalance flows hit, while the lunchtime lull barely moves the needle, so the real work lands after 3pm.

How this was measured

Using UBER minute bars over the last ~36 months, restricted to US regular-session (Mon–Fri, 09:30–16:00 ET). For each trading day with both windows present, computed: (i) the window's high–low range as a share of that day's RTH high–low range, with the midday 12:00–14:00 share divided by 2 to put it on a per-hour basis; and (ii) the window's compounded return via log-sum of minute returns, compared against the day's net open-to-close return (approximated as first-vs-last minute close). A paired t-test (ttest_rel) evaluates whether the per-hour range share is larger in the closing hour than in midday. Correlations and sign-alignment rates quantify whether the closing hour ‘decides’ the day's direction.

The key numbers

Days analyzed (both windows present)
752
2023-06-30 to 2026-06-30
Mean range share — power hour
28.9779%
Window range / daily RTH range
Median range share — power hour
26.1691%
Mean range share — midday (per-hour)
20.6737%
Midday 2h share divided by 2
Paired t-stat (per-hour share: power vs midday/2)
16.505
Positive favors power hour
Paired p-value
0.0000
p=0.0000 < 0.05 → power-hour per-hour range share > midday
Corr(power-hour ret, daily ret)
0.202
N=752 paired days
Corr(midday ret, daily ret)
0.476
N=752 paired days
Sign alignment — power hour vs daily
58.112%
Sign alignment — midday vs daily
62.899%
Days power-hour explains ≥50% of |daily move|
32.623%
Days midday explains ≥50% of |daily move|
44.607%
Per-hour share: power hour > midday share rate
74.468%
Fraction of days with pw_share > md_share/2

Reading the numbers

Across 752 days the last hour takes about 29.0% of a day’s RTH high–low range per hour versus ~20.7% for a typical midday hour — a highly significant gap (paired t=16.505, p≈1.89e-52). But midday returns correlate more with the full-day return (0.476 vs 0.202).

The charts

Per-hour share of daily high–low range
What this chart says

This box plot compares the per-hour share of the daily high–low for the closing hour versus an average midday hour. The power hour sits higher (mean 0.2898, median ~0.2617) and shows a longer right tail (max 0.809) versus the midday per-hour max of 0.5, so the end of day more often concentrates big intraday ranges. In plain terms: the volatility is thicker after 3pm, with more extreme range days coming from the close.

Mean per-hour range share by window
What this chart says

The mean-share bar chart makes the gap obvious: power hour mean = 0.2898 versus midday per-hour mean = 0.2067. That visible gap is the one the statistics flag as significant, so yes, on average the last hour takes a larger slice of the daily range than a typical midday hour. This supports the 'closing hour concentrates volatility' part of your thesis.

Power-hour return vs daily open-to-close return
What this chart says

The scatter of power-hour return against the day's open-to-close return is a low-slope cloud: power-hour returns themselves are tiny on average (x mean ≈ -0.0001) while daily returns have a wider spread (y min -0.0901 to max 0.1139). The modest positive correlation reported (0.2024) and a sign-alignment rate of about 58.1% show the closing-hour move is only a weak predictor of the full-day direction.

Direction/impact metrics: power hour vs midday
What this chart says

This comparison of directional metrics flips the narrative: midday (12:00–14:00) shows a higher correlation with the daily return (0.476 vs 0.2024), a slightly better sign-alignment rate (0.629 vs 0.5811), and more days where that window explains ≥50% of the daily move (0.4461 vs 0.3262). Put bluntly, while the power hour packs more range, the midday stretch actually lines up with and explains daily direction more often.

Window summary — range share and return-direction linkage

windowN_daysmean_range_sharemedian_range_sharemean_returncorr_with_daily_retsign_align_rateshare_>=50%_abs_move
Power hour (15:00–16:00)7520.28980.2617-0.00010.2020.5810.326
Midday (12:00–14:00)7520.41350.385700.4760.6290.446

The takeaway

Short answer: the final hour after 15:00 is measurably louder per hour, but it doesn't reliably ‘decide’ the day’s direction. On a per-hour basis the close captures about 28.98% of the day’s high–low range versus 20.67% for the midday hour, and the power hour has a higher per-hour share on roughly 74.47% of the 752 days examined; the paired t-test is t=16.51 with p≈1.89e-52, so the range-concentration result is essentially certain in this sample. That said, midday returns line up with the day’s net open-to-close return more often: correlation with the daily return is r=0.476 for midday versus r=0.202 for the close, sign-alignment is 62.9% vs 58.1%, and midday explains at least half of the day’s absolute move on 44.6% of days compared with 32.6% for the close. Bottom line: expect more raw range per hour after 3pm (useful for volatility or short-term range plays), but don’t assume the closing hour automatically determines the day’s net direction — the lunchtime window actually shows stronger directional linkage in this three-year sample.

The fine print