Gaps and session transitions

A gap is a measured change between two session references

A market gap is the difference between a new session’s opening reference and a reference from the previous session, usually the close or official settlement. If yesterday closed at 100 and today’s first eligible opening price is 103, the close-to-open gap is +3%.

That definition is more important than the empty space drawn on a chart. Futures trade across multiple sessions, equities have pre-market activity, and some instruments use an official auction or settlement that differs from the last visible trade. A strategy must state which sessions and prices it compares.

A gap is not a debt that price must repay. “Gaps always fill” confuses an event with one possible later path. Gap strategies usually test one of three ideas:

• continuation in the direction of the gap;
• partial or complete reversion toward the prior reference; or
• conditional behavior based on gap size and the earlier market state.

Each requires its own entry, target, horizon, and invalidation.

What can create a gap

News and repricing can occur while the primary venue is closed. Earnings, economic releases, global market moves, financing changes, and order imbalances can all move the next available price. Information, inventory, and the trading process can therefore contribute to an opening-price change before any gap rule is considered.

Some gaps represent durable new information. Others reflect thin overnight liquidity or opening inventory pressure that becomes easier to absorb once the main session is active. The strategy cannot know which type it saw from the gap alone. Context may help, but that context must be available before the later path reveals the answer.

Make gaps comparable

Raw point distance is rarely enough. A three-point gap is large for a twenty-dollar instrument and small for a high-priced index. Common scales are:

• percentage of the prior close;
• units of recent average true range;
• units of recent close-to-open dispersion; and
• percentile within the asset’s earlier gap distribution.

Normalization should use only prior information. A gap cannot be divided by the volatility of the session that has not happened yet.

The decisions inside a gap rule

A gap-reversion example

Illustrative example. Assume yesterday’s verified regular-session close was 100. Today’s first eligible primary-session price is 103, a +3% gap. Recent 20-day average true range is 2, so the gap is 1.5 ATR.

An illustrative reversion rule declares:

1. Only positive gaps between 1 and 2 ATR are eligible.
2. Wait for the first 15-minute bar to complete.
3. Enter short at the next event only if price closes below that first bar’s midpoint.
4. Target 101.50, half of the original gap, and exit at the session close if the target is not reached.
5. Invalidate above a threshold defined before the test.

If the opening print is 103 but the first executable strategy fill is 102.60, that is the entry. If price reaches the prior close two days later, it is not a successful fill for a same-session strategy.

A continuation version would trade in the opposite direction and needs a different rationale and exit. Reporting whichever version worked for each gap would be hindsight.

Design the experiment before labeling gaps

Begin descriptively. Group gaps by direction and predeclared normalized-size buckets. For each bucket, show the distribution of:

• return after the first tradable event;
• maximum continuation and retracement;
• time to partial and full fill;
• no-fill frequency within the declared horizon;
• opening spread and estimated slippage; and
• behavior around scheduled versus unscheduled information.

Then test continuation and reversion as separate strategies. Use the same calendar, scaling method, exposure, and costs. Preserve every eligible gap, including those that never retrace.

For equities, reconstruct corporate actions and point-in-time membership. For futures, distinguish genuine market movement from continuous-contract roll adjustments.

How gap research goes wrong

A common sample selects only gaps that later filled and then studies how they filled. That removes the very failures needed to estimate the strategy. Another credits an entry at the official open even though the rule needs the first bar’s completed information.

Results should weaken gracefully as timing becomes more realistic. If all edge vanishes when the entry is delayed to a knowable event, the strategy was trading the observation used to define it.

Important failure modes

• Stock splits and adjustment errors create artificial equity gaps.
• Futures roll construction creates non-tradable jumps.
• Official auction prices may not be available to the modeled order.
• Large information gaps can continue and dominate many small reversion gains.
• Timezone and holiday errors compare the wrong sessions.
• A fitted size bucket can turn ordinary sampling variation into a story.

Try it in Arizmic

Further reading

• Biais, Hillion, and Spatt, “Price Discovery and Learning during the Preopening Period in the Paris Bourse” (1999) — Shows that indicative pre-open prices become more informative as the opening approaches. It explains why an opening print can contain both learning and noise, and why venue design matters.
• Grant, Wolf, and Yu, “Intraday Price Reversals in the US Stock Index Futures Market: A 15-Year Study” (2005) — Directly studies intraday behavior after large opening gaps in US stock-index futures. Its filter sizes and futures setting make the evidence conditional rather than support for “gaps always fill.”
• Akbas, Boehmer, Jiang, and Koch, “Overnight Returns, Daytime Reversals, and Future Stock Returns” (2022) — Documents an asymmetric relation between positive overnight returns and negative daytime reversals in US equities and investigates competing trader clienteles. The monthly cross-sectional design is not a ready-made single-gap trading rule.