Outside Day

the mirror of the inside day — expansion exhausts, but direction stays hidden

An outside day is one where the session’s high is above the prior day’s high AND the low is below the prior day’s low — the bar fully engulfs the previous bar’s range. It is the exact inverse of the inside day. Per Brooks, an outside bar is a two-sided fight where both buyers and sellers got activated; the range blows out, and most often it is a trading-range bar — a transition rather than a clean signal. This study asks the mirror-image questions of the Inside Day study: does the blown-out range mean-revert, and does the bar’s close location carry direction into the next session? Across 1,475 outside days on the same 15 liquid US large-caps and ETFs (Jan 2022 – May 2026), the answer rhymes with the inside-day result almost perfectly.

Outside-day contraction rate
74.3%
CI95 [72.0%, 76.5%]

Next session’s range is smaller than the outside day’s range on nearly 3 in 4 outside days.

Baseline rate
48.4%
CI95 [47.6%, 49.2%]

On any random day (non-outside), next day’s range is smaller than today’s about half the time.

Edge over baseline
+25.9 pp
CI95 [+23.5, +28.3 pp] — entirely above zero

0.77× range in the next session (0.94 ATR vs 1.22 ATR outside-day mean). The blow-out reverts.

The range cycle is symmetric

Read alongside the Inside Day study, the two results close a loop. Inside days compress and resolve into expansion (72.9% expand, +26.6 pp). Outside days expandand resolve into contraction (74.3% contract, +25.9 pp). Both effects are nearly the same size and both run against the same ~48% baseline. And in BOTH cases the magnitude is predictable while the direction is a coin flip — the inside day’s midpoint carry was 49.7%, the outside day’s close-location carry is 48.2%. The market mean-reverts its range; it keeps the direction coin hidden either way.

H13 · Inside day
compression → expansion · 72.9% · dir carry 49.7%
H14 · Outside day
expansion → contraction · 74.3% · dir carry 48.2%

Outside-day vs. baseline contraction rate

After outside day
74.3%
n=1,475 · CI95 [72.0%, 76.5%]
Baseline (non-outside days)
48.4%
n=14,785 · CI95 [47.6%, 49.2%]

50% reference = coin-flip. Difference = +25.9 pp (CI95 entirely above zero). 15 tickers · 2022–2026. n=5,000 bootstrap.

What IS predictable: the range reverts

An outside day reliably signals that the next session will have a SMALLER range than the outside day itself. The contraction rate is 74.3% across 1,475 events — all 15/15 tickers show contraction rates between 69% and 81%, with the weakest name (MSFT at 69.2%) still 19 pp above its own baseline. The mean next-day range is 0.77×the outside day’s range in ATR units (0.94 vs 1.22 ATR). The blow-out exhausts: expansion → contraction is structural, the inverse of the inside-day result.

What is NOT predictable: direction

Whether the outside bar closed near its high (a “bull” outside bar) or near its low (a “bear” outside bar) does NOT predict the next session’s close direction. Close-location carry rate: 48.2% (CI95 [45.3%, 51.0%] — straddles 50%, on 1,199 decisive bars). Statistically a coin flip — and if anything fractionally below 50%, hinting the strong close is more often a trap than a tell. Same lesson as the inside day: you know the range will shrink; you don’t know which way the next session breaks.

Per-ticker detail

Sorted by outside-day contraction rate. All 15 tickers contract between 69–81%. The close-location carry column shows noise throughout — no ticker achieves a CI-clear directional edge.

TickerN outsideContract %BaselineEdgeDir carry
AMZN8681.4%47.1%+34.3 pp46.3%
QQQ10781.3%46.9%+34.4 pp47.3%
AVGO9378.5%48.2%+30.3 pp51.4%
IWM11276.8%48.5%+28.3 pp50.5%
XLK10075.0%49.2%+25.8 pp47.7%
XLF10774.8%48.0%+26.8 pp41.5%
NVDA10874.1%47.9%+26.2 pp47.3%
AAPL9173.6%47.8%+25.8 pp50.0%
JPM10672.6%48.1%+24.5 pp48.3%
SPY11372.6%49.4%+23.2 pp48.9%
AMD10072.0%47.9%+24.1 pp50.7%
GOOG9471.3%48.5%+22.8 pp51.3%
TSLA7570.7%50.1%+20.6 pp47.0%
META9269.6%49.2%+20.4 pp56.5%
MSFT9169.2%49.7%+19.5 pp39.5%

What this means for trading

  1. Outside days are range-contraction setups. After an outside day, the next session very likely offers a smaller range than the outside day itself (74.3%, mean 0.77×). Expected-move and options pricing that extrapolate the blown-out range forward will tend to overshoot — the day after an outside bar is, on average, calmer.
  2. Do NOT trade direction off the close location.The 48.2% close-location carry means a strong close near the high or low of an outside bar is coin-flip information at best. On an outside day, do not pre-position the next-day trade based on where the bar closed. Wait for the next session’s own break.
  3. Outside bars confirm Brooks’ “trading-range bar” framing.The contraction-with-no-direction signature is exactly what you’d expect from a two-sided fight that traps both sides. The outside bar resolves the volatility, not the trend — it is a balance event, not a breakout.
  4. It is the inside day in reverse. Use the two together: an inside day flags a likely range EXPANSION tomorrow; an outside day flags a likely range CONTRACTION. Neither tells you the direction. Sizing and volatility expectations key off which one just printed.

How this fits the other studies

  • Inside Day: The direct mirror. Inside days compress and expand (72.9%); outside days expand and contract (74.3%). Same baseline (~48%), same edge size (~26 pp), same coin-flip direction. Together they describe a daily range that mean-reverts around its own recent volatility — wide days beget narrow ones and vice versa — while keeping direction independent of the compression/expansion signal.
  • Gap Fill Rate: Both studies point to mean reversion at the daily scale. Small ATR-normalized gaps fill 75% of the time; outside-day ranges revert downward 74% of the time. Volatility — whether a gap or a blown-out range — tends to be given back rather than extended.
  • Day Types: An outside day is frequently the bar that ENDS a directional run and ushers in a trading range — the opposite role from the inside day, which often precedes a trend day. The Day Types classifier should see range/reversal sessions cluster after outside bars and trend sessions cluster after inside bars.

Pre-registered verdicts

H14(a) — contraction rate > 70%: PASSED. 74.3% contraction rate (CI95 [72.0%, 76.5%]). All 15 tickers above 69%.
H14(b) — CI95 of difference over baseline is entirely > 0: PASSED. Diff = +25.9 pp, CI95 [+23.5, +28.3 pp] — large and unambiguous. The outside-day reversion is real, not noise.
H14(c) — close-location carry > 52% with CI clear of 50%: FAILED. 48.2% carry (CI95 [45.3%, 51.0%]) over 1,199 decisive bars. Where the outside bar closed does not predict the next day’s close direction. Coin flip.

As with H13(c), the failure is the useful part: it rules out the “strong close = follow through” intuition and confirms the Brooks framing of the outside bar as a two-sided, non-directional event. Compression and expansion both resolve their MAGNITUDE predictably and leave DIRECTION to the next session.

Honest caveats

  • “Contraction” is partly mechanical: an outside bar engulfs the prior range, so it is a large-range bar by construction (1.22 ATR mean), and large bars are followed by smaller ones on average. The point of the study is to QUANTIFY that reversion (0.77×, 74.3%) and to test the direction question — not to claim contraction is surprising.
  • The directional test uses only decisive bars (close in the top or bottom third). A finer test — magnitude of the close location, or conditioning on the prior trend (outside bar after a sell-off vs after a rally) — could still surface a reversal edge this coarse split misses.
  • Universe is the same 15 mega/large-cap US names — all highly liquid. The effect may differ for low-float names where an outside bar can form on a single headline and carry more information.
  • Daily OHLC only. Whether the next-day contraction is tradeable depends on intraday path; an intraday study would refine when the smaller range is established.

Methodology

1,475 outside-day events + 14,785 baseline events on 15 names (2022-01-03 – 2026-05-30) via Massive/Polygon daily bars. Outside day: strict engulf (today.H > prev.H AND today.L < prev.L). Contraction: next-day range (H-L) < outside-day range. ATR normalization uses Wilder ATR(14) strictly pre-signal. Close-location carry: loc = (close − low)/(high − low); bull bar loc > 0.66, bear bar loc < 0.33 (1,199 decisive bars) → predicts next-day close vs open. Bootstrap CI (n=5,000, seed=17). Engine: scripts/ml/backtest_outside_day.py. Run 2026-06-16.

Related: Inside Day · Gap Fill Rate · Day Types