TSLA overnight gap fill rate by size (last ~3 years)
Big gaps do not behave like noise. I measured 750 TSLA overnight-gap sessions over the last ~3 years and found that the fill rate collapses with size: the largest-decile gaps filled same day only 18.67% of the time, while the smallest gaps nearly always reversed. So the folk rule “gaps always fill” is only true for noise-sized opens; the biggest gaps rarely round-trip intraday.
Below I test that thesis rigorously — minute bars limited to US regular hours, gaps defined versus prior close, fills flagged by touching the prior close, and days grouped into deciles by absolute gap size. The full analysis presents the directioned fill rates, decile breakdowns, and the statistical correlation showing a strong negative relationship between gap magnitude and same-day fill probability.
For TSLA over the past ~3 years, do overnight gaps actually 'always fill' — what share of gap-up and gap-down opens trade back through the prior close by the bell, and does that fill rate collapse as the gap widens? Thesis: modest gaps round-trip intraday the large majority of the time while top-decile gaps rarely fill same-day, so 'gaps always fill' holds only for noise-sized moves and fails exactly when the gap is big enough to matter.
How this was measured
Filtered TSLA minute bars to US regular-session (Mon–Fri, 09:30–16:00 ET). For each session, computed the overnight gap as (09:30 open − prior session close)/prior close. A gap-up is marked filled if the same-day intraday low touched or fell below the prior close; a gap-down is filled if the intraday high touched or rose above the prior close. Grouped gap days into in-sample deciles by absolute gap size and measured same-day fill rates overall and by direction to test whether fill probability declines as gaps widen.
The key numbers
Reading the numbers
Across 750 gap sessions, about 66.83% of gap-ups and 60.35% of gap-downs filled by the close; the smallest-decile gaps filled ~98.67% while the largest-decile filled only ~18.67%. Spearman ρ = -0.514 (p ≈ 1.0e-51) says larger gaps fill much less often.
The charts
This bar chart ranks same-day fill rates by absolute-gap decile: the leftmost bar (D1) is almost always filled at 98.67%, while the rightmost bar (D10) is rarely filled at 18.67%. The decline is steady but accelerates in the top deciles (D7–D10 drop into the 49.33% → 18.67% range), so the bigger the gap, the less likely it is to trade back through the prior close by the bell. For your question, that means "gaps always fill" is true for noise-sized moves (D1–D3) but breaks down dramatically for the largest gaps.
This split-by-direction chart shows the same overall pattern for both gap-ups and gap-downs: tiny gaps fill nearly every time (Gap-up D1 = 100%, Gap-down D1 = 97.22%), while the largest gaps almost never do (Gap-up D10 = 16.67%, Gap-down D10 = 20%). Note a mild asymmetry in some middle deciles—for example Gap-up D3 = 92.86% versus Gap-down D3 = 66.67%—but both directions converge to low fill rates as gap size grows. In short, modest gaps of either sign usually round-trip intraday, but top-decile gaps of either sign rarely fill same-day.
Gap-direction summary
| direction | N | fill_rate | median_gap_pct | median_abs_gap_pct |
|---|---|---|---|---|
| gap-up | 407 | 0.6683 | 0.0095 | 0.0095 |
| gap-down | 343 | 0.6035 | -0.0101 | 0.0101 |
Fill rate by absolute-gap decile (overall)
| decile | median_abs_gap_pct | N | fill_rate |
|---|---|---|---|
| D1 | 0.0007 | 75 | 0.9867 |
| D2 | 0.0028 | 75 | 0.96 |
| D3 | 0.0046 | 75 | 0.8133 |
| D4 | 0.0066 | 75 | 0.7467 |
| D5 | 0.0086 | 75 | 0.7067 |
| D6 | 0.0109 | 75 | 0.6933 |
| D7 | 0.0141 | 75 | 0.4933 |
| D8 | 0.0177 | 75 | 0.4533 |
| D9 | 0.0234 | 75 | 0.3467 |
| D10 | 0.043 | 75 | 0.1867 |
The takeaway
Short answer: no — gaps do not always fill. Over the last ~3 years, about two-thirds of gap-ups filled same day (66.83%, N=407) and about three-fifths of gap-downs did (60.35%, N=343). That said the effect collapses with size: the smallest-gap decile filled almost every time (98.67%, N=75) while the largest-gap decile filled only 18.67% (N=75). The relationship is strong and negative (Spearman rho = -0.514) with a p-value effectively zero (≈1.0e-51), so this is a robust pattern in 750 gap sessions rather than noise. Practical takeaway: the old “gaps always fill” rule works for noise-sized opens (the bottom decile median abs gap was tiny at 0.0007) but fails exactly when gaps are large enough to matter — the biggest decile rarely round-trips intraday, so treat big gaps as persistent moves, not automatic opportunities to fade.
The fine print
- Data covers 2023-07-05 to 2026-06-30 (≈750 gap sessions total).
- Decile cutoffs are in-sample; out-of-sample gap-size buckets may shift rates.
- Fill is detected on 1-minute bars; a one-tick cross between minutes can be missed.
- Only regular session (09:30–16:00 ET) is used; pre/post-market fills are excluded.