Reference price · path label · falsification

ES Gap Behavior: Definitions, Study Design, and Trade Tests

By Kyle ParrottUpdated August 25, 2026

An ES chart can show three legitimate “gaps” at the U.S. cash open: the difference from CME’s prior daily settlement, the difference from the prior 4:00 p.m. ET cash-session reference, and the difference from an overnight reference chosen by the researcher. Those gaps can disagree in size or even sign. A fill-rate claim without the reference is not a result.

Choose the reference Write the failure rule

Reference: Named and timestamped
Open: Executable rule
Path: Forward-safe labels
Fixed fill odds: Not claimed

A gap is a subtraction, not a story

Name Both Endpoints and Their Price Rules

The study must define the old reference and the new observation with exact timestamps, instruments and price construction. CME distinguishes daily settlement from final settlement, and the NYSE core cash session has its own close. None should be silently substituted for a last chart print.

Gap name	Old reference	New reference	Main question
Settlement-to-open	Prior official CME ES daily settlement	Registered cash-open execution benchmark	How far is the open from the exchange benchmark?
Cash-close-to-open	Prior 4:00 p.m. ET cash-core reference, with an explicit price rule	Same registered cash-open benchmark	How much repricing occurred between cash cores?
Globex-session gap	Prior trade-date close or session reference	Next official session’s first eligible quote/trade	What happened across the exchange pause or session boundary?
Overnight inventory gap	Research-defined overnight anchor such as VWAP	Cash-open benchmark	Where did the cash open occur relative to overnight trading?

Signed normalized gapg_t = (P_new,t − P_reference,t) / P_reference,tKeep the raw point gap too. If using volatility scaling, the denominator must be available before the new observation.

CME publishes daily settlement procedures and separately explains final settlement through Special Opening Quotations. The NYSE session guide identifies the normal 9:30 a.m.–4:00 p.m. ET cash core. These are distinct clocks and prices.

Build the eligible day before seeing the path

Construct a Timestamp-Safe Sample

Use dated ES contracts and a documented roll policy. Record source timestamps, sequence integrity, holidays, early closes and whether the cash open existed as expected. If a platform supplies a continuous chart, keep the raw contracts so adjustment artifacts can be audited.

Resolve the trade date.Map exchange time, UTC and cash-market calendar without hard-coded daylight offsets.
Choose the active contract.Use a rule based on information available then; flag migration rather than selecting the eventual volume leader with hindsight.
Capture endpoints.Store the official settlement and the quote/trade records needed to reproduce each open or close benchmark.
Run integrity gates.Reject missing sequences, stale or crossed books, unverified timestamps and incorrect instrument joins.
Freeze eligibility.Apply event, holiday and minimum-liquidity rules before labeling fill or continuation outcomes.

First print is not automatically an executable open.

A trade print records a transaction between specific counterparties. If the strategy assumes an order at the cash open, simulate its arrival, order type, latency, visible capacity and partial fills. Otherwise the study measures price paths, not attainable trades.

Label only after the required horizon ends

Define Fill, Partial Fill and Continuation Without Leakage

The decision features must be frozen at entry time. Outcome labels may use later prices, but those labels cannot be inputs to the same decision. Declare the horizon and touch rule in advance.

Full fill

Reference is reached

The eligible bid, ask or trade—whichever was preregistered—reaches the entire gap reference within the horizon. State whether a touch counts.

Partial

Path retraces but stops short

Record maximum favorable retracement as a fraction of initial gap size rather than hiding it inside a binary miss.

Continuation

Path extends away first

Measure adverse extension before any fill, with the same price side and latency rules used for execution.

Censored

Horizon ends or data fails

Keep “not observed” separate from “did not fill.” Early close, halt or feed loss can censor the outcome.

A path record that can be audited

Store gap definition, signed and absolute size, volatility-scaled size, opening spread and depth, maximum favorable and adverse excursion, first-touch time, eligible fill quantity, close-of-horizon state and censor reason. Report both directions and do not combine them until symmetry is tested.

For limit-style fill studies, a price touch is insufficient when queue and quantity matter. For marketable entries, measure the book sweep and realized shortfall. A directional backtest that enters at an untradeable print must remain a hypothetical price-path study.

The gap carries overnight information

Condition on Information and Market State

Compare outcomes inside prespecified strata rather than searching every combination for a flattering fill rate. A useful minimum set covers gap size, direction, overnight path, prior-day state, scheduled information, opening liquidity, weekday, holiday and roll regime.

Gap-scale controls

Raw points
Return percentage
Pre-open volatility units
Sign and absolute magnitude

Information controls

BLS/BEA release timestamp
FOMC communication state
Unscheduled official notices
Unknown event label retained

Execution controls

Spread and depth at decision
Order size and latency
Roll and holiday status
Rejected/censored observations

Use the official BLS, BEA and Federal Reserve calendars for scheduled labels. They establish publication timing, not the sign or duration of a price reaction.

A filled gap can still be a losing trade

Simulate the Entire Order Path and Cost Stack

Separate the descriptive question “did price revisit the reference?” from the economic question “could the declared order policy earn after costs?” The second requires entries, exits, stop behavior, latency, queue assumptions, commissions and slippage.

Element	Required assumption	Conservative treatment
Entry	Order type, side, arrival time, quantity	Cross the spread or model queue; never use an unavailable midpoint
Exit	Reference touch, target or time stop	Use executable side and available depth
Protection	Stop trigger and order behavior	Include gap-through and partial-fill risk
Fees	Commission and exchange/broker charges	Use dated all-in rates, not a zero-cost placeholder
Capacity	Size relative to displayed/replenished depth	Stress reduced depth and slower acknowledgement

Write the result that would make you stop

Use Holdout and Retirement Rules

Fit definitions and thresholds on a training period, lock them, then run one forward holdout. Report uncertainty, sample sizes and all attempted variants. A research claim fails if it depends on one reference, one subperiod or optimistic fills.

Reference instability

Settlement-gap and cash-close-gap conclusions conflict materially.

Result: narrow or reject the claim

Cost failure

Gross path behavior disappears under plausible executable costs.

Result: no tradable evidence

Holdout failure

The frozen rule fails its preregistered forward threshold.

Result: retire, do not retune

Date ranges, exclusions and observation counts are visible.
Each gap definition is reported separately.
Open, touch and exit price sides are explicit.
Partial paths and censored sessions are not forced into binary outcomes.
Event and roll strata retain enough observations for honest uncertainty.
Multiple-testing choices and abandoned variants remain in the research log.

Research status as of August 25, 2026

No original ES gap-fill probability, continuation rate or after-cost edge is reported. This page is a test protocol. Claims that gaps “usually fill” remain unverified here until a defined sample, execution model and untouched holdout pass.

Sources, methods and editorial disclosure — reviewed August 28, 2026

CME daily settlement procedures for official daily benchmarks.
CME Equity Index final settlement procedures for the separate Special Opening Quotation process.
CME E-mini S&P 500 contract page and CME trading-hours calendar for contract and schedule controls.
NYSE trading-session FAQ for the cash core-session clock.
CME historical futures and options data catalog for trade, top-of-book, depth and market-by-order distinctions.

Sources were reviewed August 28, 2026. The proposed labels and tests are editorial research design, not CME definitions or completed empirical findings.