Central limit order book · replay · execution limits

6C Order Flow: What Futures Data Can Actually Show

A 6C footprint is a record of transactions on a particular futures venue and contract. It is not a window into every bank quote, spot trade, forward, swap, option hedge, or internalized customer flow in the global Canadian-dollar market.

Venue
CME futures
Book
Displayed orders
Global FX
Partially observed
Original result
None

Venue evidence

The 6C Book Is Precise and Incomplete

Fact: CME Globex operates a central limit order book for listed futures, and CME describes its Market by Order feed as order-level data with individual queue information, full depth, and individual displayed order sizes. Fact: the BIS Triennial Survey covers a much larger global OTC FX market. Inference: 6C order-book evidence can be exact for the recorded venue while remaining incomplete as a measure of total CAD trading.

Visible on the futures venue

Trades and displayed book changes

With an entitled and correctly decoded feed, a researcher can observe exchange timestamps, prices, quantities, order additions, changes, deletions, depth, and trade events for the selected dated contract. The precise fields depend on the feed and message specification.

Not visible from 6C alone

The rest of the Canadian-dollar market

The book does not reveal bilateral OTC quotes, internalized dealer flow, every spot venue, forwards, swaps, options positions, or why a participant traded. Nor does an order ID provide a trader’s identity or economic motive.

CME’s Market by Price format consolidates quantity at price levels, while Market by Order preserves more order-level detail. A footprint made from completed trades is narrower still. State which dataset is present before discussing imbalance, absorption, queue, or cancellation behavior.

Contract identity first

Order flow belongs to a dated 6C contract. Around roll, activity can migrate to the next expiry. Verify instrument definitions and current mechanics through the canonical 6C contract guide; do not combine expiries into one synthetic book.

Fail before analysis

Reconstruct the Book or Do Not Claim Book Behavior

An OHLCV bar cannot recover the order sequence inside the bar. A trade-and-quote file cannot support queue-position claims if it omits order identifiers. A message stream with unrepaired sequence gaps can create false cancellations, false depletion, or impossible depth. Match the claim to the data actually retained.

Claim familyMinimum evidenceFail-closed condition
Trade imbalanceSequenced trades, price, quantity, reliable aggressor classification rule, synchronized quotesTrades cannot be aligned to the prevailing book or classification is ambiguous.
Depth imbalanceBook snapshots plus all applicable incremental updates for declared levelsSequence gap, stale snapshot, unsupported depth, or unresolved reset.
Cancellation rateOrder-level add/change/delete messages and lifecycle-consistent IDsFeed aggregates orders by price or an order lifecycle is incomplete.
Queue estimateOrder-level messages, exact submission acknowledgment, matching rules, latency recordTrader order or exchange queue position cannot be reconstructed.
Historical fillDecision-time book, order rule, size, latency, queue model, partial-fill logic, feesReplay uses future depth, assumes touch equals fill, or ignores queue loss.

Feed-integrity record

Store channel, instrument ID, schema version, packet and message sequence, exchange and receipt timestamps, snapshot status, recovery events, clock offset, and entitlement provenance. Preserve raw immutable input so the reconstruction can be replayed after a parser fix.

Sequence
Continuous
Snapshot
Matched
Clock
Audited
Raw input
Immutable

Observable variables

Translate the Message Stream Into Auditable Measures

Every derived variable needs a formula, sampling rule, and known limitation. A colorful visualization is not a definition.

Quoted spread

Best ask minus best bid at a declared timestamp. Store time-weighted distributions and missing-book intervals.

Displayed depth

Visible quantity at the touch or within a fixed number of price levels. It can cancel before an order arrives.

Book imbalance

A frozen transformation of bid and ask depth over specified levels. Results depend on depth horizon and sampling.

Trade imbalance

Buyer-initiated minus seller-initiated quantity under a disclosed classification rule. Ambiguous prints require an explicit policy.

Replenishment

Displayed quantity added after trades at a fixed price within a frozen time window. It does not identify hidden intent.

Price response

Midquote or executable-price change after a declared event, with several frozen horizons and the full path retained.

Normalize measures by state where appropriate. Ten contracts at the best bid can mean something different in an ordinary one-tick market than during a release with a wider spread and rapid cancellations. Preserve raw units alongside any z-score or percentile so the economic scale remains visible.

Level first, narrative second

Anchor the Hypothesis to an Objective Price Region

Order-flow labels become circular when a researcher sees a reversal, draws a level, and then calls the preceding trades absorption. Define the level without using the future response. Possible sources include a prior session extreme, an official settlement, a pre-event range boundary, or a volume reference computed entirely from past data.

1

Declare the level

Name its source, timestamp, tolerance band, and expiration rule before the first eligible touch.

2

Declare the flow event

Define the exact imbalance, trade rate, replenishment, or cancellation threshold using past and current messages only.

3

Declare outcomes

Measure forward movement, adverse excursion, time to exit, spread, and fill behavior over frozen horizons.

4

Declare the baseline

Compare with eligible level touches that lack the flow event and with clock-, event-, and volatility-matched observations.

Candidate hypothesis

Replenishment changes the response distribution

At a predeclared level, repeated displayed replenishment after marketable trades may be associated with a different forward-price distribution than matched touches without replenishment.

Prohibited inference

“A bank is defending this price”

The futures feed does not identify a bank, hedge, customer, or motive. Even an order-level pattern cannot establish a participant’s total position across venues.

Event-time validation

Replay the Decision at the Speed It Could Have Existed

Research timestamps must reproduce what a live process could know. Compute the signal only after all required messages arrive, add declared processing and order-entry latency, and evaluate the fill against later book events. Using the exchange timestamp for the signal but zero latency for the order gives the hypothetical trader information and speed it did not possess.

Replay componentFrozen ruleOptimistic shortcut to reject
Signal timeLast required input plus computation latencyTrigger at the first message that only later completes the pattern.
Arrival timeSignal time plus measured or stressed network/order latencyAssume immediate exchange arrival.
Marketable orderConsume visible quantity by price until size is filledFill the entire order at the best quote.
Passive orderTrack queue ahead, cancellations, trades, partial fills, and expiryCount every touch as a complete fill.
OutcomeUse executable prices and deduct commissions and feesScore midquote movement as realized P&L.

Partition chronologically and keep the final period untouched until definitions, parser version, exclusions, thresholds, and code are frozen. Group uncertainty by event or trading day because message-level observations within one episode are dependent.

What can imitate a pattern

Audit the Roll, News Clock, and Data Path

An apparent imbalance effect can be the signature of a scheduled announcement, a switch between expiries, a stale book, or a parser restart. Treat each as a rival explanation.

Scheduled news

Tag Bank of Canada, Statistics Canada, Federal Reserve, BLS, and BEA timestamps. Estimate ordinary and event samples separately.

Contract migration

Measure nearby and next contracts independently; never aggregate their depth as one venue.

Hidden liquidity

Displayed size is not total trading interest. Native iceberg behavior and replenishment complicate naive cancellation ratios.

Cross-venue reaction

Spot or rates can incorporate news before 6C. A 6C response does not establish that 6C discovered the information.

Message bursts

Sampling at fixed intervals can miss order and cancellation sequences. Event-time and clock-time views should agree directionally.

Weekly positioning

CFTC reports aggregate positions by category and time. They do not identify intraday aggressors or explain a footprint.

Robustness should vary adjacent level bands, flow thresholds, sampling intervals, latency, order size, and event exclusions according to a frozen grid. If the sign exists only at one narrow parameter or in one contract month, classify the result as unstable.

Ending record

Pass the Data-Quality Gate Before Reviewing the Trade

This article does not report an original order-flow result. It produced no 6C order-book reconstruction, footprint study, imbalance estimate, replay, fill result, backtest, or live-performance record. The following data-quality record is a gate for future work, not evidence that any named pattern works.

GateRequired evidenceDisposition if missing
Instrument and rollDated contract, reference definition, migration statusReject observation
Feed integrityNo unresolved sequence gap; snapshot and incrementals agreeReject interval
Clock integrityExchange/receipt clock audit and known latencyNo lead-lag or replay claim
Hypothesis provenanceLevel, signal, threshold, horizon, and baseline frozen in advanceExploratory label only
ExecutionQueue, partial fills, slippage, fees, size, and latency modeledNo profitability claim
Post-event reviewExpected and actual signal time, book state, fills, path, confounds, invalidationDo not add to validated sample
Sources, methods and editorial disclosure

Sources and methods were reviewed August 13, 2026. This article defines an original data and replay protocol but reports no original order-flow finding. It is unsponsored editorial analysis.