Does Volume Matter Before a Breakout? A Study of 11,240 Trades
Version 1.0 · report 2026-10-09 · study recorded 2026-10-08

QUESTION AND FINDING
High volume in a completed breakout bar went with better average returns in this sample. High volume the previous day gave a weaker positive clue. Neither finding establishes that waiting for high volume would improve a live trade: the strongest measurement includes trading that happened after the price crossed the entry level.

METHOD
Research question: Do volume conditions distinguish average returns among archived bullish consolidation trades, and when was each condition knowable? This is an observational association study, not a randomized experiment or a prospective live strategy test.
Population: 11,240 admitted and resolved bullish stock consolidation trades, 1,824 symbols, entry years 2016–2025. The primary archive retains its original family, admission, deduplication and lifecycle rules. It is not a point-in-time reconstruction of every security or every candidate pattern.
Entries and exits: Each row retains the archived trigger/gap entry, structural stop, target and original lifecycle exit. No hourly-close confirmation or next-bar entry is added. A 30-minute high locates the first interval reaching the saved trigger; it does not identify the exact transaction time or prove a live fill. There is no change to entry or exit between volume categories.
Outcome and price allowance: Net return (%) = 100 × [(saved exit × 0.9975) / (saved entry × 1.0025) − 1]. This applies 0.25% adverse prices on each side. The raw price return and average R are secondary outputs. The allowance is not a complete model of commissions, spreads, liquidity, market impact, taxes or account capacity.
Daily-volume definitions: Previous day: D−1 volume divided by mean volume in the 20 sessions before D−1. Recent: mean volume in the 5 sessions before entry divided by the preceding 20-session mean; disjoint windows. Consolidation trend: second-half mean divided by first-half mean before entry. Daily definitions remain those of the archive.
Categories: Low ≤0.75×; normal >0.75× and <1.5×; high ≥1.5×. The same ratio thresholds define decreasing / flat / increasing for consolidation halves. Flat denotes a broad middle band. The completed breakout interval uses the same-clock-slot historical average, not a full-day average.
Intraday reference: Massive split-adjusted 30-minute aggregates. Exchange calendar: XNYS, America/New_York, exchange_calendars 4.13.2. Seek 20 usable full sessions within 60 preceding exchange sessions. Skip shortened or incomplete historical sessions individually. Include shortened entry sessions using their actual close.
Price and symbol checks: Repository class-share aliases and provider-dated historical ticker identities are used. Prior daily/intraday closes set the median price-scale ratio; each selected prior ratio must be within 2% of that median. The factor is not tuned to force an entry-day crossing. Gaps strictly after the first crossing are allowed.
Available-before measurement: Cumulative volume in completed regular-session intervals strictly before the first crossing interval, divided by the mean volume over the identical clock window in the 20 historical sessions. This covers 4,361 later crossings and can lag the crossing by up to 30 minutes. The 6,671 opening-interval crossings have no such snapshot.
Coverage accounting: 11,240 = 11,032 verified intraday measurements + 208 unavailable. The 208 comprise 56 entirely absent entry sessions; 25 missing intervals preventing first-crossing identification; 18 daily/intraday trigger disagreements; 66 prior price-scale disagreements; and 43 with fewer than 20 usable reference sessions. All remain in the daily-volume panel.
Corrections and research sequence: The earlier hourly-close / next-open experiment was withdrawn. The corrected study retained all original entries and exits, skipped unusable historical sessions individually, and recovered historical symbols. Earlier outcomes had already been seen. An October 8 amendment fixed the corrected rules before calculation; it was not an untouched preregistration. The direct high-versus-low follow-up came after viewing the corrected charts.
Intervals around group averages: 2,000 percentile bootstrap resamples clustered by stock and separately by entry month, seed 81028. Use the lower of the 2 lower bounds and higher of the 2 upper bounds as the reported 95% envelope. This is a sensitivity envelope from 2 one-way cluster analyses, not a formal joint two-way cluster estimator; unmodeled dependence may remain.
Direct differences and multiple comparisons: 10,000 draws per clustering, seed 81027, for 10 specified follow-up contrasts. Report ordinary 95% intervals and 99.5% intervals, the Bonferroni adjustment for 10 comparisons at a 5% family allowance. This adjustment covers those 10 contrasts; it does not correct every prior research choice or convert exploration into confirmation. Intervals containing 0 are not proof of equivalence.
Monte Carlo sensitivity: The prior 4-contrast analysis used 2,000 draws and 98.75% intervals. With 10,000 draws, the known-before low-minus-normal 98.75% lower endpoint moved from approximately −0.009 to +0.083 percentage points. It still crosses 0 under the expanded 10-comparison adjustment (99.5% lower approximately −0.055). Near-zero endpoints should not be treated as stable trading edges.
Verification and reproducibility: Checks authenticated all 11,240 saved outcomes, 11,032 first crossings, 14,698 source files and 87 aggregate cells. Selected bootstrap calculations were independently reproduced within this project; this is not an external replication. Public downloads preserve exact summary precision, all 10 follow-up contrasts and source hashes. Licensed source-market responses and the full trade-level archive are not redistributed in this article.

SOURCE
Source: TradingPal’s corrected October 8, 2026 breakout-volume study and disclosed high-versus-low follow-up. Daily archived outcomes are held fixed. Group intervals use 2,000 bootstrap draws per clustering; direct follow-up comparisons use 10,000.
Download the frozen aggregate evidence and verification record (JSON): https://tradingpal.io/research/breakout-volume-study-v1.json
Massive: custom aggregate bars and empty-interval semantics: https://massive.com/docs/rest/stocks/aggregates/custom-bars
NIST: confidence intervals for differences between means: https://www.itl.nist.gov/div898/handbook/prc/section3/prc312.htm

LIMITATIONS
- The universe is selected and does not fully represent failed or delisted companies. The years were used during system development. Results may differ for an unseen period, a different universe, bearish trades or other patterns.
- The groups contain different trades. Holding each archived outcome fixed prevents an entry-rule mismatch but does not control all differences in stocks, market regimes, volatility or pattern quality. Association does not establish causation.
- The corrected study and the high-versus-low follow-up were conducted after earlier results had been inspected. The multiple-comparison adjustment is limited to 10 listed contrasts and does not eliminate selection from the wider research process.
- Completed breakout-interval volume includes trading after an intrabar crossing. The known-before snapshot omits the current partial interval and all opening-interval entries. Neither measurement is exact volume at the transaction time.
- The 208 unavailable intraday measurements may differ systematically from the available sample. The daily comparison retains them; intraday findings apply to the measured subset.
- Returns are simulated per-trade averages, including losses. They do not model a portfolio’s capital constraints, overlapping holdings or compounding. The fixed price allowance cannot establish achievable live execution.
- Separate stock and month cluster resampling accounts for some dependence but is not a complete model of dependence across both dimensions. Only 120 entry months are represented, and interval endpoints near 0 are sensitive to resampling precision.
- Combination cells are exploratory. A taller bar or an overlapping/non-overlapping pair of mean intervals alone does not establish a reliable difference. The article reports the direct contrast intervals separately.
- TradingPal performed and publishes the research and sells trading software. This report has not been independently peer reviewed.

CITATION
TradingPal Research (2026-10-09). Does Volume Matter Before a Breakout? A Study of 11,240 Trades Version 1.0. Historical study recorded 2026-10-08. https://tradingpal.io/learn/research/breakout-volume-study