NIO overnight gap-fill vs extension (RTH) — last ~3 years
The magnet-like pull of the prior close is real but muted and asymmetric: across roughly three years of minute data, NIO’s overnight gaps touched the prior RTH close only about 52.4% of the time. Contrary to the initial thesis, gap-ups filled more often (57.6%) than gap-downs (46.9%), a split that’s statistically significant (p = 0.0039).
I measured every regular-session open versus the previous close over 751 trading days, flagged intraday “fills” when price revisited the prior close, and looked at speed and continuation. Big gaps (|gap| ≥ 1%) filled noticeably less, and fills—when they occurred—were quick (median time-to-fill ~6 minutes for up-gaps, ~4 for down). The full analysis below shows the detailed counts, charts, and robustness tests supporting these conclusions.
For NIO over the past ~3 years, do overnight gaps actually fill during the regular session — does price retrace back through the prior close after gapping up or down, or do the gaps mostly persist and extend the move? Thesis: the large majority of NIO's opening gaps get at least partially filled intraday with the prior close acting as a magnet, so fading the gap has paid off more often than chasing it, and downside gaps fill more reliably than upside ones.
How this was measured
Converted NIO minute bars to New York time and filtered to regular-session minutes (Mon–Fri, 09:30–16:00 ET). For each trading day, defined the overnight gap as the 09:30 open relative to the prior day's RTH close. A 'fill' occurs intraday when price touches back to the prior close (gap-up: low ≤ prior close; gap-down: high ≥ prior close). Computed fill rate, partial-fill fraction (share of the gap retraced when not fully filled), continuation rate (close moves further in the gap direction vs open), and, when filled, minutes from the open to the first touch of the prior close. Also compared fill rates for large gaps (|gap| ≥ 1%).
The key numbers
Reading the numbers
About 52.39% of NIO opening gaps retraced to the prior close intraday; that breaks down to 57.56% for gap-ups and 46.91% for gap-downs, so up gaps filled more often than down gaps in this sample.
The charts
This bar chart lines up fill rate vs continuation by gap direction. The fill-rate bar for gap-ups is 57.56% while gap-downs fill 46.91%; the continuation bars flip: 46.68% continuation for up gaps versus 55.90% for down gaps. Read it as: gap-ups are more likely to retrace back to the prior close, whereas gap-downs are more likely to keep moving away from the prior close.
The box plot shows time-to-fill for only the filled gaps (217 up-gap fills, 167 down-gap fills). Medians are very short — 6 minutes for up fills and 4 minutes for down fills — but the means are much higher (45.70 min up, 40.80 min down) because of a long tail (maximum fills reached ~389 and ~385 minutes). That tells you most fills happen quickly after the open (so fading at the open often works intraday), but a few fills take most or all of the trading day.
The histogram summarizes 733 absolute gap sizes with a mean gap of 2.01%, a minimum near 0 and a maximum of 18.19%. Crucially, larger up gaps behave differently: up gaps >=1% fill only 45.28% of the time (N=254), below the overall up-gap fill rate of 57.56%. In short, most gaps are relatively small and often retrace, but sizable gap-ups are notably less likely to fill.
Gap summary by direction
| direction | n_gaps | mean_gap_pct | median_gap_pct | fill_rate | continuation_rate | median_time_to_fill_min | n_gaps_|gap|>=1% | fill_rate_|gap|>=1% |
|---|---|---|---|---|---|---|---|---|
| up | 377 | 0.0201 | 0.0166 | 0.5756 | 0.4668 | 6 | 254 | 0.4528 |
| down | 356 | -0.0201 | -0.0166 | 0.4691 | 0.559 | 4 | 254 | 0.3465 |
The takeaway
Short answer: gaps do often get hit intraday, but not exactly the way your thesis claimed — overall about 52.4% of NIO overnight gaps touch the prior close during the regular session, and gap-ups fill more often than gap-downs. Specifically, up-gaps filled 57.6% of the time (N=377) versus 46.9% for down-gaps (N=356), and this direction difference is statistically significant (p=0.0039 — roughly a 4-in-1,000 chance the split is luck). For larger moves (|gap| ≥ 1%) fill rates fall to ~45.3% for up-gaps and ~34.6% for down-gaps (each group N=254), so big gaps are less likely to be fully retraced. When fills happen they come fast: median time-to-fill was 6 minutes for up-gaps and 4 minutes for down-gaps, while continuation (the close moving further in the gap direction) occurred on 46.7% of up-gaps but 55.9% of down-gaps. Bottom line: fading the gap is a modest, short-lived edge overall and works more reliably for gap-ups; downside gaps are less likely to fill and more likely to extend, so fading down-gaps carries higher risk.
The fine print
- Analysis is RTH-only (09:30 open vs prior 16:00 close); pre/post-market activity is ignored.
- A 'fill' here is any intraday touch of the prior close; requiring an actual trade-through would lower reported fill rates.
- Large-gap behavior differs and is reported separately (|gap|≥1% groups are each N=254); including many tiny gaps inflates overall fill rates.
- Minute OHLC bars can miss sub-minute touches, and early-closing sessions shorten the time available to fill.