Two-screen triage · comparable basis · residual test

GBP/USD vs 6B Divergence: Reconcile Before Trading

One screen shows spot GBP/USD at 1.28000. Another shows a September 6B last trade at 1.28240. The timestamps differ by 700 milliseconds, the first number is a midpoint, the second is a last trade, and the future has a later value date. The 0.00240 gap—24 standard 6B ticks—is not yet a signal.

Illustrative onlyNot live market data
Spot midpoint1.28000
Sep 6B last1.28240
Clock mismatch700 ms
Raw gap24 ticks

Different quote fieldsBasis not removed

First response

Freeze the Screen Before Explaining It

Save enough evidence to reproduce what each system knew. A later screenshot can show revised bars, a new front contract, or a refreshed quote and erase the original discrepancy.

Exact symbol

Vendor, venue, pair orientation, 6B contract month, security ID, and whether a continuous symbol was used.

Price field

Bid, ask, midpoint, last trade, settlement, indicative quote, or chart close; include displayed precision.

Time field

Exchange event time, vendor time, receipt time, local display timezone, and bar-close convention.

Market state

Open, maintenance, holiday, halted, stale, crossed, or thin; include spread and depth when available.

Delivery terms

Spot value date, futures delivery month, days to expiry, and observed or modeled forward points.

Feed health

Sequence gaps, reconnects, conflation, delayed permissions, outlier filters, and last successful update.

The opening numbers are hypothetical

They demonstrate how a plausible-looking difference can combine value-date basis and asynchronous fields. They are not August 2026 market prices, a fair-value estimate, or an invitation to trade.

Comparable values

Put Both Prices on the Same Clock, Side, and Value-Date Basis

6B and standard GBP/USD share the U.S.-dollars-per-pound direction. The reconciliation problem is therefore not an inversion problem; it is a timestamp, quote, contract, and carry problem.

Check in orderPass conditionIf it fails
1. InstrumentGBP/USD and the intended dated 6B contract are both identifiedCorrect the symbol; discard the screen comparison
2. TimestampPrices are aligned to a declared common UTC grid within known clock errorResample; classify as asynchronous data
3. Quote sideMidpoint is compared with midpoint, or executable buy/sell sides are pairedRecompute; do not use a last-versus-bid residual
4. Data integrityBoth feeds are fresh and sequence/recovery checks passDiscard or mark unobservable; never fill missing data with the story
5. Dated basisObserved forward points or a documented term-matched model is removedClassify the raw gap as unreconciled basis
6. Friction bandResidual exceeds both markets’ spread, latency, and model-uncertainty bandClassify as ordinary non-executable noise
residual = 6B midpoint(spot midpoint + expected dated basis)

Expected basis is not a constant. CME describes FX futures pricing as spot adjusted for the short-term interest-rate differential and time to expiry. A practical implementation should prefer observable term-matched forward points when available and preserve model uncertainty. The canonical GBP/USD and 6B comparison explains the relationship in detail.

Residual taxonomy

Classify the Residual Before Assigning a Forecast

The classification is an output. Most observed differences should end here rather than become trades.

AData mismatch

Wrong symbol, stale field, clock offset, feed gap, bar artifact, or different quote sides. Correct or discard.

BContractual basis

The difference is consistent with value date, rates, forward points, and time to expiry. Monitor, do not call divergence.

CLiquidity residual

The gap lies within executable spread, depth, latency, or temporary market-fragmentation costs. Usually untradeable.

DResearch residual

A synchronized, basis-adjusted difference survives integrity and friction checks. It qualifies for testing, not automatic entry.

Store the raw gap and every adjustment separately. A pipeline that outputs only the final residual cannot show whether a result came from changed forward points, a spot-feed substitution, a futures quote update, or a model revision.

Forward-outcome research

Test Whether the Surviving Residual Predicts Anything

There are at least three distinct hypotheses: futures catches up to spot, spot catches up to futures, or both move toward a joint value. Choose one before scoring the event. “The gap closes” is not a complete trade direction.

Minimum study design

Build simultaneous quote observations for a dated 6B contract and at least one documented spot source. Estimate basis without future information. Define a residual threshold on the training sample, impose a cooldown so one episode is not counted repeatedly, and measure both legs’ forward returns on chronological validation and holdout dates.

Unit
Independent residual episode
Outcomes
Spot, futures, and basis
Controls
Event, roll, spread, depth
Economics
One- and two-leg costs
QuestionRequired outputFailure condition
Does 6B move next?Forward 6B return conditional on signed residual, with confidence intervalEffect vanishes after timestamp offsets or event controls
Does spot move next?Forward spot midpoint and executable-side changes from the same anchorOnly one delayed vendor feed appears to catch up
Does basis normalize?Residual half-life or threshold-crossing distributionConvergence is slower than the holding period or unstable near roll
Is it executable?Net two-leg P&L with spread, slippage, commission, funding, margin, and leg riskGross residual is smaller than realistic friction
Is it robust?Multiple feeds, thresholds, sampling intervals, contract months, and untouched datesSignal depends on one vendor or one fitted threshold

A lead-lag study at millisecond horizons requires clock synchronization and latency measurement at least as precise as the claimed lead. If the uncertainty band is larger than the lead, the direction of information arrival is unresolved. Control for multiple offsets and horizons; testing each one until a peak appears is data snooping.

Execution friction is part of the hypothesis

A two-leg strategy can face different counterparties, funding conventions, margin systems, holidays, and value dates. A one-leg “catch-up” strategy is exposed to the common GBP/USD move. Backtests must model the actual accessible products, not an ideal institutional spot midpoint.

Four-way decision gate

Reconcile, Discard, Investigate, or Test—Never Automatic Trade

End the workflow with the least aggressive action supported by the evidence.

Reconcile

A known quote, timestamp, contract, or basis difference explains the screen gap. Correct the display or calculation and log the cause.

Discard

Data integrity, market state, or clock uncertainty prevents a defensible comparison. Mark the episode unavailable; do not impute a price.

Investigate

A residual remains but is inside the friction or model-uncertainty band. Monitor more venues, forward points, spreads, and feed health.

Test

A clean residual exceeds frozen gates. Add it to the complete event sample and wait for out-of-sample evidence before any live rule.

Research status

No original synchronized dataset, residual distribution, threshold, lead-lag estimate, basis trade, backtest or live result is reported here. The illustrative two-screen prices are fabricated for explanation, not current observations or signals. A future claim must disclose feeds, timestamp semantics, clock error, basis construction, event counts, exclusions, search space, holdout and complete execution costs.

Sources, method and editorial disclosure

Sources and time-sensitive facts were reviewed August 13, 2026. This is original, unsponsored editorial analysis.