Trend-Day Follow-Through
the extreme gets exceeded the next day — the close does notA daily “trend bar” is a session that closes on its extreme with a large body on above-average range — a bull trend day closes in the top quartile with a body spanning at least half the bar; a bear trend day is the mirror. Per Brooks, a trend bar is the market voting: one side has clearly won, the chart is “always-in” that direction, and “there is almost always at least a small amount of follow-through the next day.” This study asks two questions of 2,811 trend days on 15 liquid US large-caps and ETFs (Jan 2022 – May 2026): does the next session push past the trend day’s extreme, and does that push survive to the close? The answer is yes to the first and no to the second.
After a bull trend day, the next session prints a higher high 3 in 4 times — vs a 52.7% baseline.
After a bear trend day, the next session prints a lower low 69% of the time — vs a 45.6% baseline. The edge is symmetric.
Next-day CLOSE direction is a coin flip (53.2% up). The follow-through is exceeded early, then reverts.
Continuation rate vs. baseline
Bull edge = +22.4 pp (CI95 [+20.1, +24.8]) · bear edge = +23.5 pp (CI95 [+20.9, +26.0]) — both entirely above zero. 15 tickers · 2022–2026 · n=5,000 bootstrap.
What IS predictable: the extreme is exceeded
A trend day reliably signals that the next session will trade beyond its extreme in the trend direction — 75.1% for bull days, 69.1% for bear days, both ~22–23 pp above their own baselines and present in all 15 names. Because the edge holds for both directions, it is momentum, not the bull-market drift of 2022–2026 (drift would inflate the bull number and deflate the bear one — instead they are nearly equal). When continuation happens, the next bar carries 0.95 ATRpast a bull high / 0.64 ATR past a bear low. This is Brooks’ “always-in” follow-through, measured.
What is NOT predictable: where it closes
The follow-through does not survive to the bell. After a bull trend day the next session closes higher only 53.2%of the time — a hair above the 53.0% all-days baseline (diff CI95 [−2.5%, +3.0%], straddles zero). The bear case is the same coin flip. So the early push past the extreme gets sold (or bought) back: at the daily scale these liquid names mean-revert into the close. You can lean on tomorrow tagging today’s extreme; you cannot lean on tomorrow closing in the trend’s favor.
Per-ticker detail
Sorted by bull continuation rate. Every one of the 15 names exceeds its prior extreme more often after a trend day than the ~50% baseline — the effect is universal, not driven by one or two names.
| Ticker | N bull | Bull → HH | N bear | Bear → LL |
|---|---|---|---|---|
| AAPL | 97 | 81.4% | 84 | 70.2% |
| SPY | 90 | 81.1% | 101 | 73.3% |
| AMD | 99 | 80.8% | 91 | 65.9% |
| TSLA | 99 | 79.8% | 100 | 63.0% |
| XLK | 96 | 79.2% | 111 | 66.7% |
| META | 84 | 78.6% | 83 | 71.1% |
| AMZN | 91 | 74.7% | 85 | 67.1% |
| NVDA | 96 | 74.0% | 90 | 71.1% |
| IWM | 88 | 72.7% | 102 | 71.6% |
| AVGO | 113 | 72.6% | 94 | 64.9% |
| QQQ | 97 | 72.2% | 108 | 66.7% |
| JPM | 86 | 72.1% | 81 | 77.8% |
| MSFT | 88 | 70.5% | 106 | 66.0% |
| XLF | 89 | 68.5% | 91 | 72.5% |
| GOOG | 85 | 67.1% | 86 | 72.1% |
What this means for trading
- A trend day is a follow-through setup for the OPEN, not a swing thesis. The 75% / 69% continuation rate says: after a strong-close day, expect the next session to trade through that extreme at some point — usually early. That is where the edge lives. Targeting the extreme tag (a measured-move continuation, a breakout-pullback-go) is supported by the data.
- Do not hold the continuation trade into the close blindly.The close-stick coin flip (53% / 47%, zero edge) means the next session frequently reverses after tagging the extreme. Take the follow-through move; don’t assume the day closes in the trend’s direction. This is the same daily mean reversion the Gap Fill study found at the open.
- The symmetry is the tell.Bull and bear edges are within 1 pp of each other (+22.4 vs +23.5 pp) across a period that trended up overall. A drift artifact would be lopsided. Equal edges in both directions is the signature of a real, mechanical follow-through tendency — exactly what Brooks describes as “always-in” pressure spilling into the next bar.
- Combine with the extension number for stops/targets.Given continuation, the next bar runs ~0.95 ATR past a bull high (0.64 ATR past a bear low). That sizes a realistic first target for a continuation entry and frames where a protective stop sits relative to the trend day’s extreme.
How this fits the other studies
- Inside Day: the mirror image. An inside day is compression that resolves into expansion but hides the direction; a trend day is expansion that resolves into directional follow-through but hides the close. Both land on the same lesson — the chart tells you THAT tomorrow moves, and often which extreme it tags, but not where it settles.
- Gap Fill Rate & First-Hour Fade: the close-stick failure is daily mean reversion, the same force that fills small gaps (75% under 0.25 ATR) and fades first-hour extremes. The early follow-through tags the extreme; then the responsive/ mean-reverting flow that the Gap and First-Hour studies isolate takes the close back.
- Day Types: a trend day on Day N raises the odds that Day N+1 opens with a continuation drive — the trend-from-the-open and breakout-and-go day types — even if the day ultimately reverts. Trend days seed directional opens the way inside days seed range expansion.
Pre-registered verdicts
The H14(c) failure is the useful half: it confirms Brooks’ specific wording — follow-through the NEXT DAY (the extreme is tagged), not a multi-day swing. The trend day’s information is spent on the open, not the close.
Honest caveats
- “Continuation” is a binary tag of the extreme (next high > trend high) — it does not require the move to be tradeable size or to come early in the session. The 0.95 / 0.64 ATR mean extension suggests most tags are meaningful, but an intraday-path study would confirm whether the push reliably happens in the first hour (tradeable) or only late.
- Trend-bar thresholds (body ≥ 50% of range, close in top/bottom quartile, range ≥ ATR(14)) are reasonable but discretionary. Stricter thresholds (e.g. close in top 10%, range ≥ 1.5 ATR) would isolate stronger trend bars at the cost of sample size; the qualitative result is robust to the exact cut, but the rates would shift.
- Universe is the same 15 mega/large-cap US names as the prior studies — all highly liquid and index-correlated, which is exactly the population where daily mean reversion is strongest. The close-stick coin flip may not hold for low-float momentum names that trend multi-day.
- Daily OHLC only. The study cannot see whether the next-day extreme was tagged before or after a reversal, nor the path between. Continuation and close-stick are end-of-bar facts, not a simulated entry/exit.
Methodology
1,398 bull + 1,413 bear trend days against 16,245 baseline day-pairs on 15 names (2022-01-03 – 2026-05-30) via Massive/Polygon daily bars. Trend bar: body ≥ 50% of range, close in the top (bull) / bottom (bear) quartile of range, and range ≥ Wilder ATR(14) measured at the bar BEFORE the signal (no lookahead). Continuation: next high > trend high (bull) / next low < trend low (bear). Close-stick: next close vs trend close. Baselines computed over all consecutive day-pairs. Bootstrap CI (n=5,000, seed=17). Engine: scripts/ml/backtest_trend_day.py. Run 2026-06-22.
Related: Inside Day · Gap Fill Rate · First-Hour Fade · Day Types