Returns · rolls · multiple testing
6A Seasonality: How to Test Calendar Effects
A colorful monthly chart is not evidence of a durable 6A seasonal edge. The result changes when you alter the contract roll, sample start, return definition or months inspected. Write the hypothesis first, test returns instead of price levels, control the search and reserve unseen data. Anything less is a story fitted to a calendar.
- Result reported
- None
- Test variable
- Returns
- Roll rule
- Documented
- Validation
- Unseen data
Define
One calendar rule
Build
Auditable series
Test
Return distribution
Hold out
Unseen years
No invented month biasCosts stay in
Direct answer
Seasonality Is a Testable Hypothesis, Not a List of Favorite Months
6A may show calendar differences in a particular sample, but this page does not claim that any month is reliably bullish or bearish. No original seasonal dataset or coefficient was produced for this article. The defensible job is to explain how to test a narrow calendar claim without letting hindsight, roll artifacts or repeated searching manufacture the result.
Valid question
Does one predefined calendar bucket differ out of sample?
Specify the return interval, contract series, roll rule, sample, benchmark and test before calculating the answer. Then report effect size and uncertainty, not just a winning percentage.
Invalid shortcut
Which month looks best on the full chart?
Scanning every month, weekday, week-of-month and holding period uses the same history to invent and verify the idea. The best-looking cell will usually overstate reality.
Standard 6A represents 100,000 AUD and quotes U.S. dollars per AUD. The current CME Globex outright tick is 0.00005, worth $5. A 0.0001 pip is two ticks, or $10. Final settlement is physical. Seasonal research must still identify the dated contract and roll because the continuous symbol is not itself a tradable expiry.
Pre-registration
Write One Falsifiable Claim Before Opening the Results
A useful template is: “Using official daily settlements and the stated roll rule, the average 6A return during calendar bucket A differs from benchmark bucket B over the training sample and remains economically meaningful in a held-out sample after costs.” The bucket might be a month, part of a month or days around a recurring event. Do not name the winning bucket after looking.
Specify standard 6A, exact expiries and whether the rule may hold through roll. Do not mix spot AUD/USD and futures silently.
Define settlement-to-settlement simple or log returns. State inclusion rules for missing or holiday observations.
Freeze month, day or event-relative boundaries and the timezone used to assign them.
Choose all other observations, a matched regime sample or another predeclared comparator.
Preselect mean difference, median, regression coefficient or another justified statistic and its uncertainty measure.
Set the minimum after-cost effect that would matter. Statistical significance alone does not pay spread or slippage.
If you want to examine twelve months, admit that you are testing twelve related hypotheses. Adding entry days, holding periods and filters multiplies the search again.
Series construction
The Roll Rule Is Part of the Result
CME's continuous price series documentation distinguishes mappings such as active and front contracts and uses official settlements. Any vendor series can apply its own mapping and adjustment. Record the methodology, retain raw dated contracts and rerun the test under plausible alternatives.
| Decision | Defensible treatment | Failure if ignored |
|---|---|---|
| Settlement field | Use the same official or verified vendor settlement definition throughout | Mixing last trade and settlement shifts returns |
| Contract selection | Define front, active or liquidity-based mapping before the test | Look-ahead can choose the winner after the fact |
| Roll date | Apply an observable volume, open-interest or fixed-calendar rule | Price jumps and volume migration contaminate the bucket |
| Adjustment | State none, difference or ratio adjustment and why | Back adjustment can alter long-horizon levels and returns |
| Timezone | Use the settlement trading date consistently | Local midnight rules move returns across months |
| Missing days | Publish holiday and missing-data handling | Unequal intervals are treated as ordinary daily returns |
The RBA publishes historical AUD exchange rates that can be a useful robustness series, but spot and futures are different instruments with different timestamps and carry. Agreement strengthens context; disagreement requires investigation rather than cherry-picking.
Testing workflow
A Seven-Step Seasonal Protocol
Keep a dated research memo and code version so another person can reproduce the output.
Freeze the question
Write the bucket, comparison, return, sample dates, roll and statistic before viewing the seasonal table.
Audit the raw data
Check duplicates, gaps, impossible returns, expiry transitions and settlement-field changes.
Calculate returns
Use the declared formula. Keep gross price movement separate from dollar profit and loss.
Describe the sample
Report count, mean, median, dispersion, tails and results by year. Plot observations, not only averages.
Control the search
Adjust for the hypotheses inspected or use a formal data-snooping procedure. Disclose every rejected variation.
Hold out data
Reserve later years or use rolling validation. Do not retune the rule whenever the holdout disappoints.
Apply costs and risk
Convert commission, spread and slippage into ticks. Stress gaps and contract-roll execution before considering use.
Robustness
Ask Whether the Calendar Is Only a Proxy
A month can line up with recurring policy meetings, commodity cycles, fiscal dates or shifts in global risk without being the cause. Compare the calendar coefficient before and after relevant controls. The goal is not to force significance; it is to learn what the label is capturing.
Macro controls
Relative rates, commodities and the U.S. dollar
The RBA identifies relative interest rates, commodity prices and risk sentiment among important AUD drivers. Use observable measures known at the time. The commodity-cycle guide explains why one commodity is not the whole export story.
Market controls
Volatility, liquidity, events and positioning
Tag major releases, roll windows and volatility states. Historical CFTC Commitments of Traders data can provide weekly positioning context, but its categories and reporting lag do not identify who caused a daily move.
Repeat the test with alternative reasonable roll dates, simple versus log returns, winsorized versus raw tails, and spot AUD/USD as a robustness check. Publish disagreement. Robustness is not the act of trying variations until one passes.
Failure modes
Most Seasonal Errors Are Research-Design Errors
The calendar story is often cleaner than the evidence.
Trending levels make unrelated variables look connected. Test returns.
One monthly observation per year creates fragile estimates and wide uncertainty.
The best of many tested rules inherits selection bias even if its isolated p-value looks impressive.
A relationship from one rate, commodity or policy era may disappear in the next.
Future volume or an undocumented stitch can embed information unavailable in real time.
A small average return can vanish after spread, commission, slippage and failed execution.
If a seasonal effect cannot survive a documented series, search correction, unseen data and costs, call it a historical pattern, not an edge. A clean negative result is better than a market myth.
Frequently asked questions
6A Seasonality Questions
Does 6A have a reliable best or worst month?
No month claim is established on this page because no original 6A seasonal dataset was run here. A reliable claim would need a documented price series, enough independent years, realistic costs, multiple-testing control and stability in unseen data.
Should 6A seasonality be tested with prices or returns?
Use returns for the statistical test. Price levels can trend and create spurious calendar relationships, while settlement-to-settlement simple or log returns provide comparable changes across different price regimes.
How should contract rolls be handled in a 6A seasonal study?
Use individual contracts or a continuous series with a fully documented roll and adjustment rule. Test nearby roll choices because a stitch, back adjustment or liquidity migration can create or erase an apparent calendar effect.
How many years are needed to test 6A seasonality?
There is no magic minimum. Twelve monthly observations require twelve years, and those observations may still share the same macro regimes. Report the count, dependence, uncertainty and performance across subperiods rather than relying on a small average.
Can a seasonal pattern be used as a standalone 6A signal?
A calendar label alone is weak evidence. If a pattern survives out-of-sample testing, treat it as one conditional input alongside current rates, risk, commodities, liquidity and execution cost, with a predefined invalidation rule.
Sources, method and editorial disclosure
- CME Group FX Product Guide 2026 for 6A specifications.
- CME Group Continuous Price Series for official-settlement and contract-mapping methodology.
- RBA historical data for official AUD exchange-rate robustness data.
- RBA Explainer: Drivers of the Australian Dollar Exchange Rate for macro control variables.
- CFTC historical Commitments of Traders data for optional positioning context.
- White (2000), A Reality Check for Data Snooping for multiple-testing risk.
No original seasonal return, month ranking, coefficient, p-value or backtested trade is reported here. The examples are research-design templates, not findings. Sources and time-sensitive facts were reviewed August 13, 2026. This is original, unsponsored editorial analysis.