AI Research COST

COST 5-day forward returns after worst-decile down-days vs baseline (last ~3 years)

Short answer: buying the Costco dip didn't prove to be a reliable edge over the last ~3 years. After days in the worst decile (daily drop ≤ −1.3158%), the mean 5‑day forward return was −0.15% versus a +0.45% unconditional baseline — a raw gap of about −0.61 percentage points. Only 45.3% of those selloffs finished positive compared with 59.1% for the baseline.

The study resampled minute bars to daily closes (≈36 months), producing 75 worst‑decile events among 746 eligible start days. The gap leans toward underperformance but is not definitive: Welch t = −1.55, two‑sided p = 0.124, and event volatility is higher. The full report below shows the charts, distributions, and robustness checks that explain where the dip idea breaks down and what tighter filters would need to show.

The research question

For COST over the past ~3 years, does 'you can't go wrong buying the Costco dip' actually pay — after its worst-decile single-day drops, do forward 5-day returns beat the everyday baseline, or does the dip just keep bleeding? Thesis: forward 5-day returns after sharp selloffs come in above baseline, so in a strong secular uptrend the drops mean-revert and dip-buying is a genuine edge rather than a comforting slogan.

How this was measured

Resampled COST minute bars to daily closes over the last ~36 months, computed close-to-close daily returns, and defined 'worst-decile' selloffs as days at or below the in-sample 10th percentile of daily return. For each eligible start day, measured the forward 5-trading-day return as close[t+5]/close[t]−1. Compared the distribution of forward 5-day returns after worst-decile drops to the unconditional baseline of all days with a t+5 close, using Welch's two-sample t-test (unequal variance).

The key numbers

Analysis window start
1688083200s
2023-06-30 to 2026-06-30
Eligible start days (t+5 exists)
746
Worst-decile threshold (daily return)
-1.3158%
10th percentile of daily returns within window
Event count (worst-decile days)
75
Mean 5d fwd — worst-decile
-0.1525%
N=75
Mean 5d fwd — baseline
0.4532%
N=746
Edge (event − baseline)
-0.6057%
Fraction positive — worst-decile
45.3333%
Fraction positive — baseline
59.1153%
Welch t-statistic
-1.554
Positive favors worst-decile bucket
Welch p-value (two-sided)
0.1238
p=0.1238 ≥ 0.05 → no statistically-clear difference

Reading the numbers

Across 75 worst-decile single-day drops from 2023-06-30 to 2026-06-30, the average 5-day return was -0.0015 versus a baseline +0.0045 — a shortfall of about -0.0061 (0.61 percentage points). The difference is not statistically clear (Welch p ≈ 0.1238).

The charts

COST 5-day forward returns: worst-decile vs baseline
What this chart says

This box plot lines up the 75 worst-decile down-days against the full baseline (N=746). Look at the centers and tails: the worst-decile group has a mean of -0.0015 while the baseline mean is +0.0045, and both share the same left extreme (min = -0.1204) while the baseline reaches a slightly higher upside (max 0.0827 vs 0.0731). The overall shift left in the worst-decile distribution — a negative mean and deeper left tail — shows that these sharp selloffs tend not to produce stronger 5-day rebounds than a typical day.

5-day forward returns after worst-decile down-days
What this chart says

The histogram shows the 75 individual 5-day outcomes after worst-decile drops spanning from -0.1204 up to 0.0731 with a mean of -0.0015. Note that fewer than half of these events were positive (fraction positive = 0.4533), so recoveries within five days are not the majority outcome. That concentration of small losses and occasional large left-tail events suggests the dip often keeps bleeding rather than reliably mean-reverting over five days.

Mean 5-day forward return: worst-decile vs baseline
What this chart says

This simple bar chart puts the two means side-by-side: -0.0015 after worst-decile drops versus +0.0045 for the baseline, a gap of about -0.0061 against dip-buying. While the raw gap favors the baseline, the Welch t = -1.554 and two-sided p = 0.1238 indicate the difference is not statistically clear, so these data do not support a confident, repeatable five-day edge to buying the Costco dip.

Forward 5-day return summary

groupNmeanmedianstdfraction_positivep5p95
Worst-decile75-0.0015-0.00090.03250.4533-0.06720.0416
Baseline7460.00450.00630.02840.5912-0.03830.048

The takeaway

Short answer: no — in this 2023-06-30 to 2026-06-30 sample, COST days that fell into the worst decile (daily drop ≤ -1.32%) produced an average 5‑day forward return of -0.15%, while the unconditional baseline averaged +0.45%. The raw gap is about -0.61 percentage points (75 worst‑decile events vs 746 baseline starts), and only 45.3% of those worst‑decile cases finished positive versus 59.1% for the baseline. The difference is suggestive but not definitive: Welch t = -1.55 with a two‑sided p = 0.124 — roughly a 12‑in‑100 chance this shortfall is noise, and the event sample is only 75 cases. Volatility is material (event 5‑day std ≈ 3.25% vs baseline ≈ 2.84%), so while the data lean toward dips underperforming the typical day, this is a lean rather than a proven edge. Practical takeaway: blind "you can't go wrong buying the Costco dip" is not supported here; if you want to capitalize on post-drop mean reversion you need tighter filters (news, timing, or intraday signals) or more data to make a confident call.

The fine print