Realized range · event labels · forward distributions
6C Volatility Patterns: Test Compression and Expansion
Suppose an illustrative 6C window has a high of 0.73240 and a low of 0.73090. Its realized high-low span is 0.00150. That subtraction is a measured description of the window. It does not say the range was unusually small, that expansion follows, or which direction price would travel.
- Label
- Past-only
- Baseline
- Clock-matched
- Outcomes
- Size, direction, cost
- Original result
- None
Direct measurement
Calculate Movement Before Naming a Pattern
Fact: a consistent price path supports several realized-movement calculations. Observation: one window can be small or large relative to a declared reference distribution. Hypothesis: a predeclared low-movement event may change the later distribution of range, direction, or execution cost. “Stored energy” is a metaphor, not a variable.
| Measure | Frozen calculation | Useful property | Limitation |
|---|---|---|---|
| High-low range | Maximum high minus minimum low in the completed window | Transparent and maps to price distance | One extreme can dominate; path is hidden |
| True-range summary | Mean or median of bar true ranges over the lookback | Includes movement across bar boundaries | Depends on bar interval and smoothing |
| Realized variance | Sum of squared fixed-interval log returns | Uses the intrawindow return path | Bad ticks and high-frequency noise matter |
| Absolute return | Absolute start-to-end log return | Simple close-to-close magnitude | Misses a large round trip |
| Path efficiency | Absolute net change divided by total absolute path change | Separates one-way movement from churn | Highly sensitive to sampling interval |
Keep price units and returns in the research data. Convert execution distances to ticks and dollars only from current verified contract mechanics on the canonical 6C tick guide. The opening prices are a hypothetical arithmetic example, not an observed 6C event.
No hindsight box
Freeze Compression and Expansion as Separate Events
A compression label must be completely knowable at its anchor timestamp. An expansion label describes a later outcome or a separate current state; it cannot be used to redraw the compression window around the move that followed.
Compression candidate
Past-only relative movement
Example definition: the declared movement measure for the last n eligible bars falls below a percentile estimated from prior clock-matched windows. Freeze n, percentile, training length, update schedule, and tie handling.
Expansion outcome
Later movement, not automatic direction
Example outcome: forward realized range over a fixed horizon relative to its matched baseline. Also measure signed return, path efficiency, adverse excursion, spread, depth, and slippage independently.
State-ratio construction
Divide the current completed-window measure by a past-only matched median or map it to a training-sample percentile. Match at least clock window, weekday, bar interval, and contract condition. Store the continuous ratio even when the event rule is binary.
- Numerator
- Current past-only window
- Reference
- Prior matched sample
- Threshold
- Predeclared
- Future data
- Excluded
A larger later range can coexist with a losing breakout, a full reversal, or poor fills. Movement, direction, and execution are three different dependent variables. A profitable rule is a fourth question that requires complete trade logic.
Relevant comparison
Match the Clock Before Calling a Window Compressed
6C activity changes through the electronic trading day and around scheduled Canadian and U.S. information. Comparing an overnight interval with the all-day distribution can classify a quiet clock rather than an unusual state.
Use the same local-purpose window with UTC and daylight-saving labels. Avoid hard-coded offsets that drift.
Match or control weekday if the event calendar and participation differ materially.
Separate ordinary windows, pre-announcement lulls, release windows, and post-event digestion using official timestamps.
Compare the short-window state within a past-only longer-run state so crisis calm is not equated with ordinary calm.
Match normal dates, roll migration, and expiry proximity. Keep dated contracts and actual books.
Require valid quotes, trades, and feed status. Zero recorded movement during an outage is missing, not compression.
Define one primary matching design before looking at outcomes. Secondary regression or weighting approaches may test sensitivity, but they must use the same past-only information set. Report balance diagnostics and the number of candidate events lost to matching.
One episode, one anchor
Construct an Event Sample That Does Not Count Quietness Repeatedly
A compression condition can remain true across many adjacent bars. Treating every bar as a fresh event inflates the sample and gives long episodes more weight. The primary design should identify the start of an episode, impose an exit rule, and define when a later episode becomes eligible.
Clean dated data
Audit contract identity, sequence gaps, duplicates, stale quotes, crossed books, bad ticks, maintenance, and holidays.
Fit the reference
Estimate clock-matched distributions on training dates only. Freeze lookback, thresholds, and minimum sample size.
Anchor episodes
Record the first timestamp the compression rule becomes true. Apply the declared exit, overlap, and cooldown rules.
Measure forward paths
At every frozen horizon, calculate range, signed return, path, adverse excursion, spread, depth, and executable costs.
Compare baselines
Use eligible matched non-events and unconditional clock-matched distributions. Report all exclusions and attrition.
Estimate uncertainty
Block by trading day or episode and report effect sizes, tails, sample counts, and intervals.
Tag Bank of Canada decisions, Statistics Canada releases, EIA petroleum publications when relevant, Federal Reserve decisions, BLS and BEA releases, regional holidays, and CME schedule changes. If later movement is concentrated around named releases, narrow the result to that event context instead of calling it a universal compression effect.
Predeclared stress
Require the Pattern to Survive Adjacent Definitions
Compression studies have many degrees of freedom: input bars, range measure, percentile, lookback, minimum duration, cooldown, forward horizon, direction rule, session, event exclusion, and roll method. Publish the complete family and reserve a chronological final holdout.
| Stress dimension | Frozen variants | Evidence against robustness |
|---|---|---|
| Movement estimator | High-low range, true-range summary, realized variance | Claim exists only under one noisy construction. |
| Threshold neighborhood | Several coarse adjacent percentiles and durations | Sign or economic scale flips after a small change. |
| Forward horizons | Short, medium, and longer windows declared in advance | Only a post-selected endpoint looks favorable. |
| Events | All dates, ordinary dates, and named event families | Generic effect is entirely an announcement effect. |
| Contract construction | Dated contracts and two ex-ante roll policies | Continuous-series splice creates the event or outcome. |
| Execution | Observed and stressed spread, slippage, latency, and fill assumptions | Gross movement cannot support a costed rule. |
| Chronology | Development, validation, final holdout, and later monitoring | Effect decays or reverses outside the fit period. |
Correct for the entire family rather than the reported winner. Report confidence intervals and economic scale, not only p-values. If event episodes overlap in their forward horizons, use dependence-aware uncertainty or thin the sample according to a predeclared rule.
From movement to execution
A Range Result Still Needs a Direction and Fill Rule
Suppose a future study finds a larger forward-range distribution after compression. That result would not choose an entry side. A breakout application would still need an objective boundary, trigger, acceptance or failure definition, order type, latency, invalidation, size, exit, and stand-aside conditions.
A symmetric range finding does not imply symmetric trading results. Quote orientation, macro events, gap behavior, spread, and path can make outcomes differ.
Use decision-time quotes, order size, latency, partial fills, commissions, exchange fees, and slippage. A high-low span is not an available capture.
This is a protocol, not a result; no original finding is claimed for 6C compression or expansion. The page produced no event sample, expansion frequency, range effect, direction result, execution result, backtest, or live-performance record. The 0.73240, 0.73090, and 0.00150 values are illustrative arithmetic only.
Ending separation
Measured Fact, Open Hypothesis, Prohibited Inference
Keep these columns separate in every report. Moving a statement to the left requires actual evidence, not stronger wording.
| Measured fact | Open hypothesis | Prohibited inference |
|---|---|---|
| A named completed window has a reproducibly calculated range. | The range is unusually low relative to a past-only matched distribution. | “Energy must release soon.” |
| The episode-anchor rule on this page is specified without using forward outcomes. | Forward movement differs from matched non-events on untouched dates. | “Compression guarantees expansion.” |
| Movement, direction, and execution cost are distinct outcomes. | A frozen directional rule may add value after a validated movement result. | “More range means a profitable breakout.” |
| Official calendars can identify scheduled-event timing. | The relationship may differ on ordinary and named event dates. | “The clock alone caused the move.” |
| This article ran no original empirical study. | The protocol is falsifiable with suitable data. | Any numerical frequency, edge, or performance claim. |
Until a study passes data quality, matching, uncertainty, multiple-testing, holdout, and cost gates, the only supported application is measurement. If the later study fails, retain the negative result and reject the pattern.
Sources, methods and editorial disclosure
- CME Group Canadian Dollar futures page for current exchange-product context; detailed mechanics remain on this site’s canonical 6C specification page.
- CME Group futures and options data overview for trade, top-of-book, and depth data context. No market dataset was bought or downloaded for the protocol.
- Andersen, Bollerslev, Diebold, and Labys, Modeling and Forecasting Realized Volatility for realized-volatility construction from intraday returns.
- Bank of Canada policy interest-rate page and Statistics Canada release calendar for Canadian event controls.
- Federal Reserve FOMC calendar, BLS release calendar, and BEA release schedule for U.S. event controls.
- U.S. Energy Information Administration petroleum-status release schedule for any predeclared oil-inventory event control.
Sources and methods were reviewed August 13, 2026. This article provides an original event-study protocol but reports no original market finding. It is unsponsored editorial analysis.