Matched maturities · synchronized changes · non-circular controls
Intraday Yield Tracking for 6S: A Lead-Lag Test
A U.S. yield and 6S can rise together for opposing stories: the yield move may reflect stronger U.S. growth, inflation risk, term premium, risk relief or a common announcement, while CHF demand can respond through a different channel. Seeing two lines move is therefore not enough. The test must declare the curve points, align their timestamps and ask whether yield changes add information beyond broad-dollar and event controls.
- Yield input
- Changes, not levels
- Horizons
- Matched in advance
- Direction
- Estimated, not assumed
- Original results
- None reported
Name the curve and transformation
Define Yields, Futures Returns and the Economic Question
“Bond yields” is not a variable. Choose the issuer, instrument or curve construction, maturity, price field, sampling interval and publication semantics. Use yield changes for co-movement tests and 6S returns over the same completed horizon.
| Variable | Frozen definition | Primary limitation |
|---|---|---|
| 6S return | Dated CHF/USD futures log return from a declared trade, midpoint or bar endpoint | Rolls and stale quotes can create artificial changes |
| U.S. curve | Named Treasury cash, futures-implied or swap point at a declared maturity | Different instruments have different clocks and microstructure |
| Swiss curve | Named Confederation or money-market series at a declared maturity | Official daily observations cannot support intraday lead-lag claims |
| Relative curve | U.S. change minus Swiss change, with matched units and horizons | A constructed spread does not identify the causal component |
| Broad USD | Independent dollar return or index change available by the endpoint | Contemporaneous control is not an advance signal |
r6St,h = ln(Pt / Pt-h)|Δst,h = ΔyUSt,h - ΔyCHt,hThe Federal Reserve H.15 release describes daily U.S. Treasury constant-maturity yields. The SNB data portal publishes downloadable Swiss interest-rate and yield series. These official daily products are appropriate only for horizons consistent with their observation and release timing.
A lead exists only on one clock
Synchronize Before Assigning Lead and Lag
Vendor bars stamped at the same minute may represent different closing rules, latencies or source markets. Preserve raw event order and define how each yield observation becomes knowable. A revised or end-of-day value cannot be projected backward into an intraday decision.
Ingest
Retain source, instrument, timestamp, timezone, precision, sequence and correction fields.
Normalize
Map all observations to UTC while preserving local and exchange trade dates.
Release
Record observation time separately from official publication or vendor-availability time.
Sample
Use only completed intervals whose inputs were available by each endpoint.
If a Swiss yield series updates daily while U.S. and 6S inputs update intraday, a repeated Swiss value means “no new observation,” not “the Swiss curve was unchanged at every minute.” Either move the study to a compatible frequency or use a licensed synchronized instrument.
Remove obvious common causes
Control the Dollar, Events, Curves and Market State
A useful yield feature must add information beyond variables that can make both yields and 6S move at the same time. Controls are selected for causal relevance before results, not added until a preferred coefficient appears.
Broad-dollar factor
Use an external, timestamp-compatible USD measure to test whether the apparent yield relationship is a common dollar move.
Official events
Tag FOMC, U.S. data, SNB and Swiss releases with authoritative timestamps and separate event from non-event estimates.
Curve shape
Include a limited registered set of level, slope or policy-horizon changes so one maturity does not proxy for another.
Liquidity state
Control spreads, quote age, session, holidays and roll status where the source fields permit.
Prediction must precede the return
Run a Narrow, Preregistered Lead-Lag Grid
Estimate contemporaneous association separately from predictive lead. A feature measured through time t may predict only returns beginning after the source was observable, processing finished and a plausible order could arrive.
| Test | Feature endpoint | 6S outcome | Interpretation boundary |
|---|---|---|---|
| Contemporaneous | Same completed interval | Same completed interval | Association only; not a tradable lead |
| Short lead | Yield change through t | 6S return after declared processing delay | Must survive bid/ask and latency |
| Longer lead | Yield change through t | One of a few registered later horizons | Overlapping outcomes require dependence-aware inference |
| Reverse lead | 6S return through t | Later yield change | Tests whether direction may run the other way |
| Placebo | Time-shifted or unrelated feature | Registered 6S horizon | Detects clock and search artifacts |
Primary protocol question
Does a declared U.S., Swiss or relative yield change add stable out-of-sample information about a later 6S return after broad-dollar, event and market-state controls, data latency and conservative execution costs?
- Baseline
- No yield feature
- Search
- Narrow registered grid
- Allowed winner
- None
A catalog of seductive errors
Reject False Confirmation Before It Becomes a Signal
These patterns can produce clean-looking overlays without incremental predictive evidence.
Clock error
Close matched to intraday
A daily official yield is aligned to an earlier 6S bar even though it was not yet available.
Common shock
One release moves both
The yield and 6S react together; neither led the other.
Circular control
FX-derived feature predicts FX
The candidate embeds CHF or a near-equivalent 6S return inside its construction.
Search artifact
Best maturity after inspection
Many curve points and lags are tried, but only the strongest historical pair is disclosed.
- Require an availability timestamp. Observation time alone is insufficient.
- Show control sensitivity. Report the estimate before and after each registered factor.
- Keep both directions. A reverse-lead test is part of identification, not an optional appendix.
- Retain failed lags. Every searched maturity and horizon stays in the multiple-testing family.
Incremental or unusable
Yield-Tracking Acceptance Record
The final record separates a descriptive relationship from a feature that can enter a live decision.
Decision: accept, restrict, or reject
- Manifest
- All instruments, maturities, sources, timestamps, transformations, lags, controls and sample hashes.
- Coverage
- Missing updates, stale intervals, event states, roll states and eligible sample counts.
- Increment
- Out-of-sample improvement versus the no-yield baseline with uncertainty.
- Accept
- Later 6S information remains stable, non-circular and useful after delay and costs.
- Restrict
- Use is confined to a state observed in real time and supported by enough independent events.
- Reject
- Relationship is contemporaneous, control-dependent, unstable, too small, too costly or data-incomplete.
No original 6S intraday yield study, coefficient, lead-lag result, predictive model or trading performance is reported here. No original result is reported. This article defines a test. Until synchronized licensed data pass the record, yields are context—not a validated 6S signal.
Sources, methods and editorial disclosure — reviewed August 21, 2026
- Federal Reserve H.15 Selected Interest Rates for official daily U.S. government and Federal Reserve rate-series definitions and publication notes.
- Swiss National Bank data portal for downloadable Swiss interest-rate, yield and foreign-exchange series.
- Federal Reserve H.10 dollar-index summary for official broad, advanced-economy and emerging-economy dollar-index definitions.
- Federal Reserve FOMC calendars and official meeting materials for U.S. policy-event timestamps and documents.
- CME Swiss Franc futures contract page for current product context; full mechanics and arithmetic belong to the canonical 6S contract guide.
- CME Rulebook Chapter 254: Swiss Franc/U.S. Dollar futures for exchange contract identity.
- CME DataMine historical-data catalog for exchange historical dataset distinctions. No data was purchased or analyzed for this article.
Sources and methods were reviewed August 21, 2026. This unsponsored article distinguishes official series definitions, proposed synchronization, hypotheses, inference and trading application. It reports no original market result.