Rule families · neutral baselines · neither is allowed
6N Trading Strategies: Compare Trend and Mean Reversion
The same 6N price path can be called a trend by a 100-bar rule and an overextension by a 10-bar rule. The chart has not changed; the lookback and task have. “Trend” and “mean reversion” become testable only after every decision and timing rule is fixed.
- Families
- 2 declared
- Baselines
- Risk-matched
- Costs
- Order-level
- Original test
- None
Comparison before optimization
Ask Which Frozen Rule Adds Decision Value
A fair comparison fixes the information set, instrument universe, bar clock, risk budget, position sizing, execution model, sample partitions and outcome metric across both families. Only the signal logic should differ.
Protocol question—not a result
Does a preregistered 6N continuation rule or a preregistered reversion rule improve a risk-matched neutral baseline in later chronological data after conservative execution costs, parameter-search disclosure and regime stress?
- Fact
- 6N is a dated future
- Hypotheses
- Trend and reversion candidates
- Allowed winner
- Neither
Persistent information or positioning may produce continuation over a defined horizon.
Temporary dislocation or liquidity pressure may reverse over a defined horizon.
Neither mechanism establishes that a particular rule predicts 6N or survives implementation.
Fully specified candidates
Translate Both Families Into Timestamped State Machines
Each candidate needs the same level of detail: eligibility, feature, threshold, confirmation, earliest order, side, size, invalidation, exit, re-entry, event policy and roll handling. Parameter values below are categories, not recommendations.
Family A
Continuation protocol
- Feature: past-only return, channel or slope over a registered horizon
- Entry: objective break or state confirmation after a completed bar
- Invalidation: price/time rule known before entry
- Exit: fixed horizon, trailing rule or opposing state chosen in advance
- Failure: late entries, false breaks and reversal gaps remain in the ledger
Family B
Reversion protocol
- Feature: standardized displacement from a past-only anchor
- Entry: objective extension plus a registered stabilization or reversal trigger
- Invalidation: displacement, price or time rule known before entry
- Exit: anchor, partial reversion or fixed horizon chosen in advance
- Failure: regime shifts, gaps and anchors that keep moving remain in the ledger
Eligible
Data, contract, event and liquidity gates pass.
Observed
All frozen features are available.
Signaled
One family’s exact condition is true.
Submitted
Order arrives after declared delay.
Resolved
Fill, miss, exit and costs are recorded.
“Avoid ugly setups,” “wait for confirmation” and “take profit when momentum fades” are untestable unless ugly, confirmation, profit and fade have exact past-only definitions.
Difficulty must be earned
Require Both Families to Beat Neutral Alternatives
A strategy can look attractive in isolation while merely reproducing market exposure, volatility targeting or trade timing. Use several simple comparisons with the same risk and executable-cost conventions.
Baseline 1
No trade
Zero market exposure and no transaction cost. A complex rule must justify taking risk at all.
Baseline 2
Direction-neutral timing
Preserve eligible times and holding periods but randomize or balance side without future data.
Baseline 3
Naive sign rule
Use a single predeclared past-return sign or anchor rule with the same sizing and exits.
Baseline 4
Risk-matched exposure
Match realized or forecast risk using only information available at each decision time.
Inference rule: outperforming a baseline in development is not confirmation. Any claim requires later untouched data and the full set of strategies and parameters tried.
Executable dated history
Use Dated 6N Contracts and Preserve the Roll Decision
CME’s rules define delivery-month futures, not an eternal continuous symbol. Keep each expiry intact, select the active contract with an ex-ante rule and separate the analytical splice from the close-and-reopen execution.
| Layer | Required evidence | Reject when |
|---|---|---|
| Contract | CME product, delivery month, quote and trade timestamps | Order cannot be mapped to an expiry |
| Bars | Timezone, boundaries, source rows, missing-data policy | Vendor bars cannot be reconstructed |
| Roll | Selection rule, decision time, fills and separate adjustment view | Future volume chooses the historical active contract |
| Features | Lookbacks, warm-up, availability lag and code version | Unfinished or future values enter the decision |
| Orders | Decision, arrival, order type, size, fill, cancel and exit | Close price is assumed executable |
CME Rulebook Chapter 258 is the primary contract authority; current mechanics and arithmetic belong in 6N Contract Specifications. CME DataMine lists settlements, Market by Order and PCAP as different data products. No dataset was purchased or tested for this article.
Chronology over shuffled rows
Develop, Select and Validate in Time Order
Random folds can leak persistent regimes and overlapping outcomes. Use expanding or rolling walk-forward partitions with purge or embargo rules long enough to separate feature windows, labels and open trades.
Develop
Define both families, limited grids, baselines, metrics, data cleaning and costs.
Walk forward
Fit only on past observations; evaluate the next block without refitting to its outcomes.
Select once
Apply one predeclared family/parameter selection rule and seal the result.
Open holdout
Assess the locked strategy once; any change requires a new untouched sample.
- Publish every candidate. Include failed lookbacks, thresholds and family variants.
- Separate trades across folds. A position cannot start in training and finish in validation unnoticed.
- Version external inputs. Revised economic data must not appear before its historical publication.
- Keep a holdout access log. Curiosity opens count as access.
From signal to fill
Charge Each Family for the Orders It Creates
Trend and reversion rules can differ in turnover, urgency and adverse selection. A flat cost per trade can favor the family whose hardest fills are hidden. Replay each declared order against evidence the feed actually supports.
side × (exit fill − entry fill) × 100,000 NZD × contract quantity−USD fees and non-embedded USD costs=Net strategy outcome in USDSpread and depth
Task-specific entry
Marketable orders walk the available book; passive orders need queue-aware evidence and retain misses.
Latency and gaps
Decision is not arrival
Apply processing and network delay, then resolve stops and fast markets beyond their trigger price.
Turnover stress
Raise costs together
Widen spread, slippage, latency and missed-fill assumptions. Reject a winner dependent on the optimistic point.
Do not select a story after failure
Predeclare Regime Tests and Rejection Rules
Continuation and reversion may plausibly respond differently to volatility, event and liquidity conditions. Regime labels must be observable at the decision time and registered before the validation outcomes are opened.
Time
Stability
- Early and late blocks
- Adjacent lookbacks
- Alternate bar boundaries
Market state
Conditionality
- Predeclared volatility states
- Event and non-event windows
- Roll and ordinary periods
Implementation
Survival
- Wider cost schedule
- Delayed arrival
- Missed and partial fills
The CFTC’s trading-system advisory warns that simulated results can benefit from hindsight and do not reproduce actual trading conditions or a trader’s ability to bear losses.
Selection verdict
Strategy-Selection Record
The final record compares both families and every baseline under one policy. The decision is selected only if all hard gates pass; otherwise record neither.
| Gate | Trend family | Reversion family | Fail-closed rule |
|---|---|---|---|
| Specification | Complete or fail | Complete or fail | Any discretionary field rejects family |
| Baseline value | Later net comparison | Later net comparison | No material improvement rejects family |
| Uncertainty | Block-aware interval | Block-aware interval | Decision threshold not resolved rejects family |
| Robustness | Full registered matrix | Full registered matrix | Reasonable alternate history reverses claim |
| Execution | Stressed net result | Stressed net result | Optimistic fills required rejects family |
| Selection | Trend, reversion, limited/experimental, or neither | Missing evidence means neither | |
No original 6N trend test, mean-reversion test, parameter search, walk-forward result, performance statistic, edge or recommendation is reported. This page is a comparison protocol, not a finding. Current selection: neither 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 exchange rules.
- CME DataMine historical-data catalog for distinctions among settlements and order-level datasets. No dataset was purchased or analyzed for this article.
- NIST/SEMATECH time-series analysis guidance for temporal dependence and chronological analysis.
- CFTC advisory on commodity trading systems and hypothetical performance for simulation, fill and hindsight limitations.
Sources and methods were reviewed August 20, 2026. This unsponsored article separates exchange facts, candidate mechanisms, rule-family hypotheses, empirical inferences that require a study and trading applications that require later execution evidence.