Distributions · event decomposition · transition uncertainty
6N Volatility Patterns: Measure Regimes Without Predicting Them
One average can combine many quiet 6N intervals with a few abrupt jumps and describe neither. Volatility analysis begins with the distribution: which horizon, which clock, which contract, which event state and which estimator created the number?
- Output
- Distribution
- States
- Past-only
- Transitions
- Probabilistic
- Original results
- None
Name the quantity
Choose an Estimator That Matches the Decision
Close-to-close variance, intraday range, realized variation and option-implied measures answer different questions. Declare the horizon, annualization, sampling, missing-data treatment and intended use before comparing values.
Return-based
Close-to-close
Standard deviation of completed returns over a frozen lookback. It compresses the intraday path and depends on the close definition.
Range-based
High–low family
Uses intrabar range under estimator-specific assumptions. Bad ticks and session boundaries can dominate the estimate.
Intraday
Realized variation
Aggregates high-frequency returns. Sampling frequency trades microstructure noise against missed movement.
Tail
Jump and quantile measures
Describe extremes or discontinuities under a declared rule. They need enough independent events and honest uncertainty.
σt,w = sqrt(sum of squared completed returns)|window, sampling and scaling declaredA measured state describes a past window. Turning it into position size, stop distance or a direction forecast is a separate rule that requires its own later validation and market-impact checks.
Source evidence before estimates
Preserve Dated Contracts, Quotes and Clock Semantics
Volatility can change when the source switches from trades to mids, the bar boundary moves, a stale quote repeats, or a continuous series rolls. The data manifest must make every observation reconstructible.
| Layer | Required fields | Failure condition |
|---|---|---|
| Instrument | 6N product, delivery month, venue and trade date | Continuous symbol has no expiry mapping |
| Market records | Exchange timestamp, price or bid/ask, quantity, sequence and corrections | Gap in sequence treated as a quiet interval |
| Bars | Timezone, exact endpoints, source-field selection and missing-bar policy | Vendor default is undocumented |
| Events | Official publication time, timezone, revision/vintage and event identifier | Modern calendar projected over history |
| Transformations | Cleaning, sampling, roll, scaling, estimator and code version | Manual removal has no audit log |
CME DataMine distinguishes settlements from order-level and packet-capture products. The source appropriate for a daily close estimator may be inadequate for intraday jump or executable-risk claims. No CME dataset was purchased or analyzed for this page.
Do not pool unlike information states
Decompose Session, Event and Ordinary Variation
A clock-time profile can proxy for scheduled information rather than a permanent property of 6N. Define bins in UTC, retain local source times and compare registered event windows with matched non-event windows.
RBNZ decisions
Official timeStats NZ releases
Versioned eventU.S. releases
Eastern timeOrdinary windows
Fixed UTC binsBuild the event table from RBNZ release guidance, the Stats NZ release calendar, and relevant U.S. agency calendars. Official schedules are inputs to a study, not evidence that an event caused a measured move.
Remove artificial jumps without hiding risk
Separate Contract Migration From Price Variation
CME Rule 258 defines a dated, deliverable NZD/USD future. A generic series may jump when the active expiry changes even if neither contract made that move. The analytical treatment and executable roll remain separate records.
Measure expiries
Compute returns and market quality within each dated contract.
Select ex ante
Use only information known before the roll decision to identify the active expiry.
Flag boundaries
Store migration dates, overlap, volume/open-interest evidence and any splice adjustment.
Stress execution
Charge both legs and model the volatility of residual exposure during the roll.
Analytical view
Return-preserving series
May support distribution analysis if every value maps to a dated contract and adjustments are disclosed.
Trading view
Actual close and reopen
Uses executable prices, spread, slippage, fees, leg timing and the chosen expiry. An adjustment is not a fill.
CME Rulebook Chapter 258 is the primary contract authority; complete mechanics stay in 6N Contract Specifications.
Classification after estimation
Define Volatility States With Past-Only Thresholds
“Low,” “normal” and “high” are not properties until the estimator, lookback and threshold rule are explicit. Calculate thresholds from the training history available at each timestamp; never use full-sample quantiles to label the past.
Eligible
Data, contract and clock records pass.
Estimate
Frozen estimator uses completed past data.
Threshold
Past-only reference distribution sets cutoffs.
Classify
State and uncertainty are stored at time t.
Act or observe
Separate validated policy decides; state alone does not.
Insufficient history produces ineligible, not a shorter improvised lookback.
If confirmation requires multiple bars, store the delay and do not backdate the state.
Near-threshold estimates may be ambiguous; the permitted action can remain no trade.
This article publishes no current or historical percentile, daily range, event move, regime threshold or timing forecast. Those values require a named sample and complete reproducible calculation.
Persistence is not prophecy
Estimate State Transitions With Honest Denominators
Volatility clustering is an empirical proposition about conditional distributions, not a guarantee that today’s state continues tomorrow. A transition table must show counts, uncertainty, duration and how states near thresholds were handled.
| Record | Required output | Misleading shortcut |
|---|---|---|
| Origin state | Past-only label, start time, duration and uncertainty | Full-sample hindsight label |
| Destination | Next fixed horizon or first transition under a declared rule | Choose horizon after seeing persistence |
| Counts | Eligible transitions, censored cases and excluded reasons | Percentages without denominators |
| Context | Event, roll, session and liquidity tags selected in advance | Explain transition with whichever story fits |
| Uncertainty | Block-aware interval or resampling distribution | Rows treated as independent when windows overlap |
Description
A separately run study may estimate how often a registered state persisted or changed in its sample.
Prediction
Using that estimate at a future decision requires a later calibrated forecast test against simple baselines.
Risk input
Turning a forecast into size or stops requires loss, liquidity, gap and execution stress.
Direction
Volatility measures magnitude, not whether 6N will rise or fall. Direction needs separate evidence.
Alternate measurement histories
Try to Destroy the Regime Claim
Choose a small, preregistered set of plausible alternatives. Robustness does not mean changing the estimator until the preferred state path returns.
Estimator
Measurement check
- Return and range family
- Adjacent sampling frequencies
- Alternate bad-tick policy
- Past-only threshold variants
Calendar
State check
- Event versus ordinary windows
- Holiday and daylight-saving variants
- Roll versus non-roll periods
- Early versus late samples
Decision
Usefulness check
- Later sealed period
- Naive persistence baseline
- Uncertainty calibration
- Execution and risk stress
NIST/SEMATECH time-series guidance treats serial dependence as part of the analysis. Report full distributions and sample attrition rather than treating every overlapping window as independent evidence.
Monitoring output
Volatility Monitoring Specification
A monitor describes the data and current past-only state; it does not forecast direction or promise that the state will persist. Missing required evidence must stop the output.
Required fields at each monitoring timestamp
- Data
- Dated contract, source, schema, clock, latest complete record and integrity flags.
- Estimator
- Name, return field, horizon, sampling, lookback, scaling and code hash.
- State
- Past-only reference distribution, thresholds, label, distance and uncertainty.
- Context
- Session, official event, holiday, roll, expiry and market-quality tags.
- Transition
- Historical conditional estimate, denominator and interval—clearly labeled descriptive.
- Action
- Observe, reduce, wait or reject only under a separately validated risk policy.
No original 6N volatility sample, typical range, event effect, state threshold, transition probability, forecast or trading result is reported. This is a monitoring protocol, not a finding. Until a reproducible study passes, status is not tested.
Sources, methods and editorial disclosure — reviewed August 20, 2026
- CME Rulebook Chapter 258: New Zealand Dollar/U.S. Dollar futures for dated-contract identity and delivery context.
- CME DataMine historical-data catalog for distinctions among settlement, order-level and packet-capture datasets. No dataset was purchased or analyzed for this article.
- CME Group trading-hours and holiday schedules for venue-state and calendar controls.
- Reserve Bank of New Zealand release guidance, RBNZ policy decision dates, and the Stats NZ release calendar for event-time controls.
- RBNZ B1 exchange rates and Trade Weighted Index for official daily NZD reference data; it is not executable CME 6N market data.
- NIST/SEMATECH time-series analysis guidance for temporal-dependence context.
Sources and methods were reviewed August 20, 2026. This unsponsored article separates contract and calendar facts, measurement definitions, candidate state hypotheses, empirical inferences requiring a study and later trading applications.