1 · Certified data

Statistics are computed only from certified historical datasets whose files are hash-verified against a data manifest, with a documented defect register and quality rules. HTA uses multiple certified research panels rather than a single fixed window; each study page identifies its exact certified source and eligible period. All clock times are shown in U.S. Eastern Time (ET), America/New_York, and are daylight-saving aware. Studies disclose three numbers separately: raw archive coverage, the certified panel used, and the study-eligible sample after exclusions.

Session-boundary studies use a back-adjusted continuous panel deliberately: raw contract series would read every contract roll as a false gap. Back-adjusted levels preserve continuity for gap, range, and timing arithmetic; they are not historical tick prices, and no continuous price series is ever published.

2 · Frozen definitions

Session windows (fixed CME/ET boundaries keyed to the CME trading date), metrics, eligibility rules, and bucket thresholds are fixed before results are computed. Two claim tiers are distinguished honestly: traceable (every published statistic ties back to a frozen definition, a certified dataset, deterministic code, and a version history) and prospectively frozen (new studies whose specification was frozen before the final computation ran). Studies state which tier they belong to.

3 · Uncertainty by default

Every published rate carries its numerator, denominator, and a Wilson 95% confidence interval. Study families apply Benjamini-Hochberg false-discovery-rate control (q = 0.05). Sample classes are labeled: OK (≥384), Thin (≥100), Sparse (≥30). Buckets below minimum thresholds never appear in headline claims. A percentage without a denominator is a slogan, not a finding.

4 · Evidence grades

5 · The publication boundary

HTA publishes research and education. Never market data. Public artifacts contain only static, aggregated, non-reconstructable findings: frequencies, medians, quantiles, intervals, and effect sizes. No session-by-session records, no downloadable datasets, no historical query tools, no continuous price series, and no live data displays are ever published. Every artifact passes a reconstruction test before release: could a reasonable user recover the underlying dataset, or use this instead of licensed market data? If the answer is anything but a clear no, it does not ship.

6 · Reviewed release

Before publication, every study passes automated research QA (sample-count drift, duplicate and missing sessions, impossible price relationships, roll anomalies, timezone checks, chart-to-table consistency, manifest validation), a rights-safe publication preflight, and recorded founder approval. Restatements are published openly with version links. Corrections are part of the method, not an embarrassment.

Library versions

Dataset windows

Restatements

One recorded restatement so far. The Wave-6 midpoint-recapture finding was restated in Wave 7: its original causal (rotation-edge) interpretation was superseded after a geometry-falsification test showed the effect was a distance artifact. The corrected study carries a restatement banner and links the prior figure and the cause on its own page. When any published figure changes, because of a data correction, definition fix, or window refresh, the old figure, new figure, and cause are recorded here and on the study page.

The research question

After NQ opened away from the prior session's regular-hours close, how often did price trade back to that close (“fill the gap”), and how did the size of the gap, measured against the prior day's range, change that rate?

Why this question matters

“Gaps usually fill” is one of the most repeated claims in trading education. If it is true only for certain gap sizes, then traders who apply it uniformly are preparing with a statistic that does not describe the situation in front of them. Knowing the size-conditional rates helps a trader frame the morning realistically instead of anchoring on a folk statistic.

What was tested

For every eligible session, the overnight gap was measured as the distance between the new session's open and the prior regular-hours close, normalized by the prior regular-hours range. A gap counted as “filled” when price traded back to the prior close. Fill rates were measured cumulatively at the end of each successive session window (Asia, London, pre-NY, regular hours), and continuation was measured as the close finishing beyond the open in the gap direction. Rates were then computed within gap-size buckets.

Eligible sample and period

2,643 eligible sessions from 2016-01-03 through 2026-06-02. The median gap was small. 0.046× the prior regular-hours range, which is itself a finding: most sessions open near the prior close. Larger-gap buckets shrink quickly (642 sessions at ≥0.10×, 198 at ≥0.25×, 73 at ≥0.50×) and are labeled with their sample class below.

Main findings

Gap direction itself was balanced: 51.2% of gaps were upward [49.2%, 53.1%], not significantly different from a coin flip. The median gap was 0.046× the prior regular-hours range.

Conditions where the pattern strengthened

The smaller the gap, the stronger the fill tendency: sessions below the 0.10× threshold (roughly three-quarters of all sessions) drove the pooled fill rate above 60%. Much of the filling happened before New York hours at all. 68.0% of all gaps had already been filled by the end of the London session, meaning the overnight tape did a large share of the work.

Conditions where the pattern weakened or reversed

Size reversed the conclusion. At ≥0.25× the prior range, filling (42.4%) was less common than continuing through the close (61.1%). At ≥0.50×, fills dropped to about a third, though that bucket holds only 73 sessions and its interval is wide [24.4%, 45.7%]. Gaps of a full prior-day range or more occurred just 14 times in a decade; no claim is published for that bucket.

Stability and sensitivity

The large-gap continuation tendency survived multiple-testing control (BH-FDR, q = 0.05) on all adequately-sampled buckets. The finding is structural (about distance and mean-reversion to a reference price) rather than directional, which historically made this family of findings more robust across market regimes than direction-based ones. A dedicated sub-period re-run (2016–2020 vs 2021–2026) is scheduled before the grade can be considered for A.

Using this research responsibly

Limitations

• Large-gap buckets are thin (n = 198 at ≥0.25×) to sparse (n = 73 at ≥0.50×); their intervals are wide and marked accordingly.
• The panel is a back-adjusted continuous series. Chosen deliberately so contract rolls do not appear as false gaps. Absolute price levels are therefore continuity-preserving, not historical tick prices; ratios and fill events are unaffected.
• “Filled” means the prior close was traded, not that a fade trade would have been profitable. No costs, execution, or path are modeled.
• The 2016–2026 window contains a predominantly rising market; while gap-fill mechanics are structural, the sub-period re-run remains scheduled.

Methodology summary

Sessions follow fixed CME/ET boundaries keyed to the CME trading date. Gap = open minus prior regular-hours close, normalized by the prior regular-hours range. Every rate carries its numerator, denominator, Wilson 95% interval, and Benjamini-Hochberg FDR flag; buckets below minimum sample thresholds are labeled Thin/Sparse and excluded from headline claims. Full methodology →

Related studies

The research question

After price first broke a completed session range, how often did it return inside the range rather than continue in the break direction?

Why this question matters

“It broke the level” is one of the most common reasons traders give for acting. But if the first break usually reverts, then treating every break as the start of a move means being repeatedly caught by noise. Knowing the revert rate reframes the break as a question. Acceptance or rejection? Rather than an answer.

What was tested

For each session, the study identified the first time regular-hours price traded beyond a completed prior range (the London range, whose boundaries are fixed by the time RTH opens). It then measured whether price returned inside that range within 5, 15, and 30 minutes, and separately how often the opposite side was ever reached. The same break-and-return test was run on the overnight ranges (Globex into Asia, Asia into London) for comparison.

Eligible sample and period

2,088 regular-hours sessions in which price broke the completed London range, drawn from 2,654 total RTH sessions, 2016-01-03 through 2026-06-02. The overnight comparison tests draw on 2,594–2,629 sessions each.

Main findings

Context: on a fifth of days RTH never left the London range at all, and it broke both sides on only 5.4%, the large overnight range frequently contained the entire New York session.

Conditions where the pattern strengthened

The reversion tendency was strongest exactly where traders are most tempted, the first break during regular hours. Overnight ranges showed the same character: first breaks of the Globex and Asia ranges returned inside within 30 minutes on 75–80% of sessions. Across every completed range tested, the first break was more often rejected than accepted.

Conditions where the pattern weakened or reversed

A returned break is not the same as a reversed one. “Back inside” describes rotation, not a profitable fade, and on the minority of days when the break held, it often led the session's real move. The finding identifies which outcome was more common, not which one will happen on any given day. Sessions that broke and accepted were the exception, but they were also frequently the trend days.

Stability and sensitivity

All three timing rates cleared multiple-testing control with large samples and are structural rather than directional, which makes them among the most regime-robust findings in the library. The failed-first-break tendency is a well-documented feature of auction markets and did not depend on the 2016–2026 uptrend.

Using this research responsibly

Limitations

• “Returned inside” measures price location, not trade profitability; path and costs are not modeled.
• The held-break minority (roughly one session in six by 30 minutes) disproportionately contains directional trend days, the tail matters more than its frequency suggests.
• Ranges are defined by fixed session boundaries; a different range definition could shift the exact rates.
• Back-adjusted panel used so rolls do not create phantom breaks.

Methodology summary

A break is the first RTH trade beyond a completed prior-session range; return-inside is measured at 5, 15, and 30-minute horizons. Every rate carries numerator, denominator, Wilson 95% interval, and BH-FDR flag. Full methodology →

Related studies

The research question

After an unusually quiet session, how often did the next same-window session expand rather than stay compressed?

Why this question matters

A string of quiet sessions lulls traders into small expectations right before the market wakes up. If compression systematically precedes expansion, then the calmest tape is often the worst moment to assume more of the same. Knowing the expansion rate helps a trader stay prepared for movement instead of being surprised by it.

What was tested

Each session's range was ranked against its trailing distribution. When a session fell in the bottom quartile (unusually narrow), the study measured how often the next same-window session's range expanded, both simply larger, and by defined multiples. The test was repeated independently across Asia, London, RTH, and the full overnight window to check consistency.

Eligible sample and period

666 bottom-quartile RTH instances feed the headline rate, part of 2,653 RTH sessions from 2016-01-03 through 2026-06-02. Each session-window version draws on 667–1,381 instances, and all four windows are computed separately.

Main findings

This is the same volatility-reversion mechanism seen on the daily chart (see the related study), observed here from the compressed side and confirmed session by session.

Conditions where the pattern strengthened

Consistency is the headline. Every window. Asia, London, RTH, full overnight. Expanded after compression at 60–68%, and the below-median variant was even stronger than the strict bottom-quartile cut. A relationship that repeats across independent session windows is far more trustworthy than one that appears in a single slice.

Conditions where the pattern weakened or reversed

Expansion says nothing about direction, and it is a tendency, not a rule. Roughly a third of compressed sessions stayed quiet. Compression can persist for several sessions before it resolves. The finding tells you the odds shifted toward movement, not that the next session must be large.

Stability and sensitivity

Every rate cleared multiple-testing control (BH-FDR, q = 0.05), and the effect appears in four independently computed session windows. Like its daily-chart counterpart, it is a structural volatility property rather than a directional one, which is why it carries the library's top stability grade.

Using this research responsibly

Limitations

• Direction is not addressed; expansion can resolve either way.
• Compression sometimes persists across multiple sessions before expanding.
• Quartile thresholds are frozen but definition-dependent; a different compression cut could move the exact rate.
• Back-adjusted panel used so rolls do not distort range measurement.

Methodology summary

Session ranges are ranked against trailing distributions with frozen quartile thresholds; expansion is measured as the next same-window range exceeding the prior. Every rate carries numerator, denominator, Wilson 95% interval, and BH-FDR flag. Full methodology →

Related studies

The research question

Which times of day historically carried the most movement and volume, and how stable was that pattern across separate test periods?

Why this question matters

Execution, event timing, and risk all depend on when the market actually moves. If you place a passive entry into the quietest block or expect a trend at lunch, the clock is working against you. Knowing the normal shape of the day makes preparation and timing far more realistic.

What was tested

Every five-minute block of the day was indexed against the average block for range, volume, and absolute return, then re-measured separately in discovery, validation, and holdout periods to confirm the shape was stable rather than a fluke of one era.

Eligible sample and period

2,155 strict-quality days from 2010-06-07 through 2026-03-06. Each block's index is normalized so a typical block equals 1.0; the values below use the most conservative reading across the three test periods.

Main findings

Volume is even more concentrated than range: the opening block ran about 30x a typical block, and the close is the second-busiest volume window of the day.

Conditions where the pattern strengthened

The shape reproduced in every test period, which is why it earns the top grade. Stable activity nodes include the 18:00 reopen, the European and macro handoffs, the 09:30 cash open, the opening hour, and the closing auction approach.

Conditions where the pattern weakened or reversed

The clock forecasts expected movement and liquidity, not direction. A high-activity block is a time to expect a wider range and to plan execution, not a reason to be long or short. And event days can override the normal shape, so live volatility still comes first.

Stability and sensitivity

Because the indices were built and confirmed separately in discovery, validation, and holdout windows, the pattern is a genuine structural feature of the trading day rather than an artifact of one period.

Using this research responsibly

Limitations

• Indices are normalized so a typical block equals 1x; they describe relative, not absolute, activity.
• Event days (FOMC, NFP, CPI) can override the normal shape and are studied separately.
• Bar volume cannot separate trade count, size, and true liquidity.
• This is descriptive seasonality, not a trade rule.

Methodology summary

Each 5-minute block is indexed against the average block for range, volume, and absolute return, then re-estimated independently in discovery, validation, and holdout periods. Published values use the most conservative reading across the three. Full methodology →

Related studies

The research question

How often did one overnight session break the completed range of the session before it, and once broken, did the break actually hold?

Why this question matters

Traders act on breaks. If the break itself happens almost every session but only holds half the time, then treating every break as the start of a move means getting chopped up by noise. The real signal is acceptance, not the break.

What was tested

For each overnight session-to-session transition, the study found the first break of the prior completed range, then measured whether it was accepted at the session close, how often price returned inside within 5, 15, and 30 minutes, and when the median break occurred. It repeated this across all three periods.

Eligible sample and period

2,229 strict days from 2014-10-13 through 2026-03-06. The headline Asia transition drew 2,076 break events; the Asia-to-London and London-to-pre-NY transitions were measured the same way.

Main findings

Because acceptance sits near 50%, both generic breakout rules and generic fade rules failed the strategy pipeline. The event worth studying is acceptance versus rejection, not the break.

Conditions where the pattern strengthened

The pattern was consistent across every overnight transition and all three test periods. It mirrors the regular-hours finding that first breaks of a completed range usually come back: the first poke through a level is more often rejected than accepted.

Conditions where the pattern weakened or reversed

A returned break is not a profitable fade, and the minority of breaks that held often led the session's real move. This describes which outcome was more common, not what will happen on any given break. Both directions of the obvious trade failed out-of-sample.

Stability and sensitivity

Break, acceptance, and return-inside rates reproduced across discovery, validation, and holdout with large samples. The near-coin-flip acceptance is exactly why generic overnight breakout and fade systems did not survive costs.

Using this research responsibly

Limitations

• Acceptance and return measure price location, not trade profitability.
• The held-break minority disproportionately contains the real trend sessions.
• Ranges use fixed session boundaries; a different definition could shift rates.
• Both breakout and fade implementations failed out-of-sample after cost.

Methodology summary

The first break of each prior completed session range is timestamped; acceptance is the session close finishing beyond the level, and return-inside is measured at 5, 15, and 30 minutes. Rates reported across all three periods. Full methodology →

Related studies

The question

How often did NQ break its opening range, and how often did the break direction still hold at the close?

What we found

What it means

A break by itself carried little information. It happened most days. The question worth studying was whether price accepted beyond the level.

The honest catch

Down-breaks held to the close only 45.5% of the time, an asymmetry that partly reflects the 2016–2026 rising market. Graded C until a bear/range sub-period re-run confirms it.

Using this research responsibly

Where it breaks

• These are historical frequencies under frozen definitions, not forecasts.
• The study shows what happened, not why, and markets change.
• Read the honest catch above, it states the main sensitivity.

Related studies

The question

What were the typical range, direction tendency, and closing behavior of each NQ session window?

What we found

What it means

Baseline context: know what a normal session looked like before judging today as unusual.

The honest catch

All upward direction tendencies in this period are partly the 2016–2026 rising market. Directional cells are graded C until re-tested on falling and sideways sub-periods; structural cells (ranges, timing) are more robust.

Using this research responsibly

Where it breaks

• These are historical frequencies under frozen definitions, not forecasts.
• The study shows what happened, not why, and markets change.
• Read the honest catch above, it states the main sensitivity.

Related studies

The question

Did scheduled macro events change how far NQ moved, which way it moved, or both?

What we found

What it means

Historical basis for treating event days as risk-sizing days rather than direction-prediction days.

The honest catch

Event samples are small (82–122 days) and graded accordingly. No directional tendency survived multiple-testing control on either event.

Using this research responsibly

Where it breaks

• These are historical frequencies under frozen definitions, not forecasts.
• The study shows what happened, not why, and markets change.
• Read the honest catch above, it states the main sensitivity.

Related studies

The question

What percentage of the time did NQ trend versus move sideways, and did that change by timeframe?

What we found

What it means

Expect rotation as the default on short intraday timeframes. Continuation is the exception, so systems that assume it are working against the base rate.

The honest catch

A descriptive habitat baseline, not an entry. It describes how often each state occurred, not when one begins. Daily trend occupancy was regime-sensitive under this exact threshold.

Using this research responsibly

Where it breaks

• These are historical frequencies under frozen definitions, not forecasts.
• The study shows what happened, not why, and markets change.
• Read the honest catch above, it states the main sensitivity.

Related studies

The question

Did Monday's gap fill more often than other weekdays once gap size was accounted for?

What we found

What it means

Drop the Monday gap-fill rule. It does not exist once you control for how big the gap was. Size is the real variable.

The honest catch

Published as an honest negative that fails a popular claim. Larger gaps filled less often and stayed on the gap side more, so condition gaps by size and volatility state, never by the day of the week.

Using this research responsibly

Where it breaks

• These are historical frequencies under frozen definitions, not forecasts.
• The study shows what happened, not why, and markets change.
• Read the honest catch above, it states the main sensitivity.

Related studies

The question

During an established intraday trend, how deep was a normal pullback before the trend resumed?

What we found

What it means

Normal trend pullbacks ran roughly a quarter of the prior leg and sometimes half. Use that to frame expectations, not to place a stop.

The honest catch

A path prior, not a stop rule. A strategy's real pullback tolerance must be tested with its own entry geometry and costs. The 4-hour and daily samples were thin and more definition-sensitive.

Using this research responsibly

Where it breaks

• These are historical frequencies under frozen definitions, not forecasts.
• The study shows what happened, not why, and markets change.
• Read the honest catch above, it states the main sensitivity.

Related studies

The question

After an unusually large first hour, how much of the day's range was already built and how much remained?

What we found

What it means

A big first hour lowers the relative amount left but rarely ends the day. Expect the pace to slow, not the movement to stop.

The honest catch

Not an entry and not a 'range is done' signal. Meaningful absolute movement typically remained even after a very large first hour.

Using this research responsibly

Where it breaks

• These are historical frequencies under frozen definitions, not forecasts.
• The study shows what happened, not why, and markets change.
• Read the honest catch above, it states the main sensitivity.

Related studies

A

Ablation

Removing one ingredient from a model to see whether it actually mattered. If the score barely moves, that ingredient was adding nothing.

Acceptance

Price closing beyond a level and staying there, not just poking through. We usually require two closes.

Asia session

The overnight window we bucket around 20:00–02:00 ET.

AUC

Area under the curve. A 0-to-1 score for how well a model separates two outcomes. 0.50 is a coin flip; higher is better discrimination, though not the same as profit.

B

Balance / mature balance

A well-formed, overlapping, low-efficiency sideways area. We tested it and found it no stronger than plain compression.

Baseline / coin flip

The “no edge” number a result has to beat, often 50%.

Beta / beta-contamination

The market’s long-run drift (up, over 2016–2026). A result that only works because the market rose is beta-contaminated, not a real edge.

BH-FDR / q-value

A multiple-testing correction. Test enough ideas and some look good by luck; the q-value is the false-discovery-adjusted p-value that guards against that.

Body ratio

How much of a bar is body versus wick, one way to describe a candle’s shape.

Boundary

The top or bottom edge of a range or box.

Boundary extension

How far price pushes a developing high or low further in the same direction.

Break / breakout

Price trading beyond a level, such as a range boundary.

C

CANDIDATE_PENDING_APPROVAL

A finding that passed the research pipeline but has not been signed off by a founder. Nothing publishes without that approval.

Candlestick families / named patterns

The popular single- and multi-bar shapes (dojis, hammers, engulfings, and the rest). Our census found no stable directional edge in any of them.

Certified dataset

A hash-verified, roll-checked data snapshot, so every number can be reproduced.

Checkpoint

A fixed clock time we freeze and measure from (for example 10:30 ET), using only the information available by then, never hindsight.

Compression

A quiet, tight range, low relative to its own recent history.

Confidence threshold

The minimum model confidence required before we count a prediction, for example 0.60.

Continuous contract

A single stitched price series built from the individual expiring futures contracts.

Contract roll / back-adjusted

Futures expire, so we splice contracts together and price-adjust the seam. A back-adjusted panel keeps rolls from faking moves.

Coverage

For a selective model, the share of days it is confident enough to make a call. Higher accuracy on fewer days is the trade-off.

D

Day types

Our five frozen labels for how a session auctioned: sustained acceptance, two-sided rotation, one-sided trend, failed auction, and mixed.

Discovery / Validation / Holdout

Three separate slices of history. We find an idea in Discovery, then it has to repeat, with the same sign, in Validation and the untouched Holdout before we trust it.

Displacement

A decisive move out of a compressed box, at least 0.10 of the box range beyond its edge.

Distance-neutral / geometry baseline

A check that strips out a mechanical, distance-based explanation to see whether a real edge is left.

E

Engulfing bar

A bar whose range swallows the prior bar’s. Another popular pattern we put to the test.

ES

E-mini S&P 500 futures. Not live in the library yet.

ET

U.S. Eastern time. Every clock time in the library is ET, in 24-hour format.

Expansion

The opposite of compression: a wider-than-usual range.

Exploratory

An evidence grade: interesting but not established. Kept out of headline claims.

F

Fail-closed exclusion

When a session is missing raw data, we drop it rather than guess, so a gap never fakes a number.

Frozen box

Any reference range whose high, low, and midpoint are locked once formed (OR, IB, prior-day, overnight, or a rolling compressed box).

G

Gap

The distance between yesterday’s close and today’s open. A “gap fill” is price trading back to yesterday’s close.

GC

Gold futures (COMEX). The second instrument in the library.

Globex / overnight

The electronic futures session that runs while New York is closed, roughly 18:00 ET to the next 09:30 ET open.

Grade A

Established. Large sample, stable across periods, definitions, and volatility.

Grade B

Established, with one meaningful caveat stated up front.

Grade C

Conditional. Valid only under defined conditions, regime-sensitive, or a smaller sample.

Gradient-boosted model (HGB)

A common machine-learning model (histogram gradient boosting) we use for day-type classification. Powerful, but scored strictly walk-forward.

H

HAR (Heterogeneous Autoregressive)

A standard volatility-forecasting model that blends recent daily, weekly, and monthly volatility. We use it as the benchmark for how forecastable range really is.

HOD / LOD

High of Day and Low of Day, the session’s final top and bottom.

I

Initial Balance (IB / IB60)

The high and low of the first 60 minutes of RTH.

Inside bar

A bar whose whole range fits inside the prior bar. We tested the popular breakout idea and it did not hold.

L

London session

The overnight window we bucket around 02:00–08:00 ET.

M

Mann-Whitney test

A distribution-free test for whether two groups differ, used when we compare, say, event-day range to a normal day.

Matched controls

A look-alike comparison group, used to check whether an effect is special or just ordinary behavior.

MFE / MAE

Maximum Favorable and Maximum Adverse Excursion: the best and the worst a trade went before it resolved.

Midpoint

The exact middle of a frozen box.

Midpoint recapture

Price displacing out of a box, then closing back across that box’s midpoint.

MNQ

Micro E-mini Nasdaq-100. Same index, one-tenth the size, so the behavior is the same but the risk per point is smaller.

Mother bar

The larger prior bar that an inside bar sits inside.

N

NO-GO

A trade idea that was tested and failed. We log the failure on purpose, so it is not re-litigated.

NQ

E-mini Nasdaq-100 futures. The main contract this library studies.

O

Opening Range (OR / OR30)

The high and low of the first 30 minutes of RTH. The day’s first reference box.

Overnight inventory

Where price sits at the RTH open relative to the overnight range, a read on who is long or short into the day.

P

p-value

The chance a result this strong could show up by luck if there were no real effect. Smaller is stronger; we still correct it for multiple tests (see BH-FDR).

Path efficiency

How directly price traveled from one point to another. Low efficiency means a choppy, rotational path.

Percentile (Q25 / Q75)

Where a value sits in its own history. Q75 means wider than 75% of comparable days.

Pre-open

The minutes just before the 09:30 ET cash open.

Prior-day range (PRIOR_RTH)

Yesterday’s RTH high and low, used as reference levels for today.

Q

Quartile

One of four equal slices of a sorted set. The top quartile is the highest 25%, the bottom quartile the lowest 25%.

R

R

A unit of the box or developed range (for example 0.25R is a quarter of that range). It lets us compare days of different sizes.

Realized volatility

Volatility actually observed in the data over a window, as opposed to implied or forecast volatility.

Rebreak

A level that gets broken, rejected, then attacked again.

Restated

An evidence grade: a prior figure’s interpretation changed. The old and new versions stay linked, with the reason explained.

Rotation / two-sided

A day that trades both boundaries instead of trending one way.

Round-trip cost

The commission-plus-slippage we subtract to keep results realistic, usually one point.

Router / habitat

HTA’s internal deployment layer. “Router frozen, 0 contracts” means nothing was traded on a study.

RTH (Regular Trading Hours)

The New York cash-equity session, 09:30–16:00 ET. Most of the index studies live here.

S

Sample size (n)

How many events a number is based on. Bigger samples give steadier numbers.

Session handoff

How one session’s range and levels carry into the next, for example overnight into RTH.

SHADOW

Decision-support only. It must run forward in a shadow period before it could ever influence a trade.

Shock (first-hour shock)

An opening hour with an unusually large range or move. We split it by direction, because an up-shock and a down-shock stamp in opposite sides of the day.

Streak

A run of consecutive same-direction closes or bars.

Sweep & reclaim

A quick spike past a level that immediately reverses back inside.

T

Target-before-stop

Whether price reached the profit target before the stop. A high rate is not the same as making money once costs are in.

V

Volatility clustering

The tendency for calm to follow calm and wild to follow wild, so volatility comes in runs.

Volatility state / regime

A label for the current volatility environment (for example calm, normal, or elevated), used as context, not a signal.

W

Walk-forward

Testing on rolling, forward-in-time slices, never on data the model already saw.

WATCHLIST

A near-miss worth re-checking later, not a trade.

Wilcoxon test

A paired distribution-free test, used when each event has a matched control.

Wilson 95% interval (CI)

The range a percentage would plausibly fall in given the sample. Wider means less certain.