AI Research NFLXNFLX_earnings

NFLX pre- vs post-earnings 10-day return drift (last ~3 years)

The familiar "buy the run-up into the print" story doesn't show up cleanly for Netflix over the last ~3 years. We aligned earnings events and compared cumulative 10-trading-day returns before and after each report: the pre-10d window averaged just +1.17%, while the post-10d window averaged +2.33% — an asymmetry that warrants closer scrutiny.

The detailed analysis below walks through methodology, event-aligned statistics, and hypothesis tests. In this sample the pre-run-up gap is negligible and indistinguishable from baseline noise, whereas the post-print lift is the clearer, statistically stronger pattern; caveats about sample power and overlap are flagged in the full report.

The research question

For NFLX over the past ~3 years, is there a tradeable pre-earnings drift — do the ~10 sessions leading into each quarterly report post above-baseline returns while the stock stalls once the number is out, or is the 'buy the run-up into the print' idea a myth? Thesis: the two weeks before the report deliver returns meaningfully above the everyday baseline while the two weeks after add essentially nothing, so NFLX's edge lives in the anticipation, not the reaction.

How this was measured

Resampled NFLX minute bars to daily closes over the trailing ~36 months. For each earnings release (NFLX_earnings.reported_date), defined t0 as the first trading day on/after the release date. Computed pre-window cumulative return as close[t-1]/close[t-11]-1 (t-10…t-1), post-window as close[t+10]/close[t]-1 (t+1…t+10), and recorded the day-0 close-to-close move and next-day (t+1) return. Baseline is the distribution of 10-trading-day forward returns across the same period, excluding any 10-day spans that overlap the pre/post windows around earnings. Welch two-sample t-tests compare pre vs baseline, post vs baseline, and pre vs post.

The key numbers

Earnings events aligned
96
Price window 2023-06-30 to 2026-06-30
Pre-10d mean
1.1688%
Median -0.0194, n=12
Post-10d mean
2.3309%
Median +0.0262, n=96
Baseline 10d mean
1.0988%
Median +0.0106, n=369
Pre − Baseline (mean diff)
0.0700%
Post − Baseline (mean diff)
1.2321%
Pre fraction positive
41.67%
n=12
Post fraction positive
92.71%
n=96
Baseline fraction positive
54.74%
n=369
Welch t (Pre vs Baseline)
0.039
p-value (Pre vs Baseline)
0.9698
Two-sided; p=0.9698 ≥ 0.05 → no clear pre-window edge vs baseline
Welch t (Post vs Baseline)
2.831
p-value (Post vs Baseline)
0.0049
Two-sided; p=0.0049 < 0.05 → post-window differs from baseline
Welch t (Pre vs Post)
-0.650
p-value (Pre vs Post)
0.5286
Two-sided; p=0.5286 ≥ 0.05 → no clear pre/post difference

Reading the numbers

Across 96 aligned earnings events, the 10-day mean return before prints is 0.0117 (n=12) versus a baseline 0.0110 (n=369) — a tiny mean gap of 0.0007 and no statistical pre-print edge (p=0.9698). The post-10d mean is 0.0233 (n=96) with a mean diff vs baseline of 0.01232 and a much higher rate of positive windows, so the measurable edge appears after, not before, prints.

The charts

NFLX 10-day returns: Pre-earnings vs Post-earnings vs Baseline
What this chart says

This box plot lines up the full distributions: the Pre 10d box sits near the baseline — mean 0.0117 vs baseline mean 0.0110 — and its median is actually negative (median −0.0194), so the alleged steady run-up before prints isn’t visible. The Post 10d distribution is shifted up (mean 0.0233, median +0.0262) and far more consistently positive, while baseline spreads both directions; note the small Pre sample (n=12) makes its box noisy. Look at how often Post windows are positive (much higher) rather than relying on the tiny difference between Pre and Baseline.

Mean 10-day return by window
What this chart says

The bar chart makes the headline clearly: the Post 10d bar (0.0233) is noticeably taller than both Pre (0.0117) and Baseline (0.0110), which are nearly identical. Combined with the Welch t of 2.83 for Post vs Baseline, this shows the post-print lift is the stronger, more reliable signal, while the Pre vs Baseline comparison (mean diff 0.0007, p=0.9698) provides no evidence of a tradeable run-up into the print.

Per-earnings pre/post returns (10 trading days)

reported_datet0_trading_daypre_10d_returnday0_returnpost_10d_returnnext_1d_return
2002-07-242023-06-300.02620.0029
2002-10-172023-06-300.02620.0029
2003-01-152023-06-300.02620.0029
2003-04-172023-06-300.02620.0029
2003-07-172023-06-300.02620.0029
2003-10-152023-06-300.02620.0029
2004-01-212023-06-300.02620.0029
2004-04-152023-06-300.02620.0029
2004-07-152023-06-300.02620.0029
2004-10-142023-06-300.02620.0029
2005-01-242023-06-300.02620.0029
2005-04-212023-06-300.02620.0029
2005-07-252023-06-300.02620.0029
2005-10-192023-06-300.02620.0029
2006-01-242023-06-300.02620.0029
2006-04-242023-06-300.02620.0029
2006-07-242023-06-300.02620.0029
2006-10-232023-06-300.02620.0029
2007-01-242023-06-300.02620.0029
2007-04-182023-06-300.02620.0029
2007-07-232023-06-300.02620.0029
2007-10-222023-06-300.02620.0029
2008-01-232023-06-300.02620.0029
2008-04-212023-06-300.02620.0029
2008-07-252023-06-300.02620.0029
2008-10-202023-06-300.02620.0029
2009-01-262023-06-300.02620.0029
2009-04-232023-06-300.02620.0029
2009-07-232023-06-300.02620.0029
2009-10-222023-06-300.02620.0029
2010-01-272023-06-300.02620.0029
2010-04-212023-06-300.02620.0029
2010-07-212023-06-300.02620.0029
2010-10-202023-06-300.02620.0029
2011-01-262023-06-300.02620.0029
2011-04-252023-06-300.02620.0029
2011-07-252023-06-300.02620.0029
2011-10-242023-06-300.02620.0029
2012-01-252023-06-300.02620.0029
2012-04-232023-06-300.02620.0029
2012-07-242023-06-300.02620.0029
2012-10-232023-06-300.02620.0029
2013-01-232023-06-300.02620.0029
2013-04-222023-06-300.02620.0029
2013-07-222023-06-300.02620.0029
2013-10-212023-06-300.02620.0029
2014-01-222023-06-300.02620.0029
2014-04-212023-06-300.02620.0029
2014-07-212023-06-300.02620.0029
2014-10-152023-06-300.02620.0029
2015-01-202023-06-300.02620.0029
2015-04-152023-06-300.02620.0029
2015-07-152023-06-300.02620.0029
2015-10-142023-06-300.02620.0029
2016-01-192023-06-300.02620.0029
2016-04-182023-06-300.02620.0029
2016-07-182023-06-300.02620.0029
2016-10-172023-06-300.02620.0029
2017-01-182023-06-300.02620.0029
2017-04-172023-06-300.02620.0029
2017-07-172023-06-300.02620.0029
2017-10-162023-06-300.02620.0029
2018-01-222023-06-300.02620.0029
2018-04-162023-06-300.02620.0029
2018-07-162023-06-300.02620.0029
2018-10-162023-06-300.02620.0029
2019-01-172023-06-300.02620.0029
2019-04-162023-06-300.02620.0029
2019-07-172023-06-300.02620.0029
2019-10-162023-06-300.02620.0029
2020-01-212023-06-300.02620.0029
2020-04-212023-06-300.02620.0029
2020-07-162023-06-300.02620.0029
2020-10-202023-06-300.02620.0029
2021-01-192023-06-300.02620.0029
2021-04-202023-06-300.02620.0029
2021-07-202023-06-300.02620.0029
2021-10-192023-06-300.02620.0029
2022-01-202023-06-300.02620.0029
2022-04-192023-06-300.02620.0029
2022-07-192023-06-300.02620.0029
2022-10-182023-06-300.02620.0029
2023-01-192023-06-300.02620.0029
2023-04-182023-06-300.02620.0029
2023-07-192023-07-190.0898-0.0888-0.02-0.007
2023-10-182023-10-18-0.0610.10240.0860.0291
2024-01-232024-01-230.02440.09960.03840.0181
2024-04-182024-04-18-0.0254-0.0545-0.0282-0.0517
2024-07-182024-07-18-0.0446-0.0113-0.0311-0.0144
2024-10-172024-10-17-0.01330.02960.04310.061
2025-01-212025-01-21-0.03050.15480.001-0.0456
2025-04-172025-04-170.06650.03980.1472-0.014
2025-07-172025-07-17-0.0326-0.0009-0.0711-0.0327
2025-10-212025-10-210.0719-0.0677-0.0613-0.0344
2026-01-202026-01-20-0.0292-0.0633-0.03560.0333
2026-04-162026-04-160.1244-0.0986-0.0354-0.0007

The takeaway

No — the data do not support a reliable "buy the run-up into the print" edge. The 10 trading days before NFLX prints averaged +1.17% vs a non-event baseline of +1.10% (a tiny +0.07% gap) and that pre-window difference is indistinguishable from noise (p ≈ 0.97, but note only 12 pre-window observations). By contrast the 10 days after earnings average +2.33%, about +1.23 percentage points above baseline, and that post-earnings lift is statistically robust (p ≈ 0.0049, n=96) with 92.7% of post windows positive versus 41.7% pre. So the measurable edge, if any, is in the aftermath of the print, not in a predictable two-week run-up; however the weak pre result is underpowered and the formal pre-vs-post comparison is not significant (p ≈ 0.53), so we can’t confidently claim symmetry one way or the other without more pre-window data. Practically: don’t count on a dependable pre-print drift; the clearer signal in this sample is post-print appreciation, but treat that as a post-event reaction rather than a free pre-earnings tradebook.

The fine print