Overnight and intraday return decomposition splits a market's move across the closed-market interval and the trading session. That simple accounting can expose patterns hidden by a daily close-to-close chart, but it does not prove that either interval is a tradable edge.
What the split measures
For a cash equity, define the overnight return from the prior session's close to the next session's open. Define the intraday return from that open to the same session's close. Compounded together, those intervals reconstruct the close-to-close move, subject to consistent prices and corporate-action handling.
The labels depend on the instrument and its venue. A nearly continuous futures or FX market has no single universal overnight boundary. State the exchange, session clock, price field and treatment of holidays before comparing results.
Why researchers separate the clock
The decomposition can distinguish a move that accumulates while a cash venue is shut from one that develops after its open. Berkman, Koch, Tuttle and Zhang report overnight gains followed by intraday reversals in their stock sample, concentrated in high-attention stocks; they discuss opening prices and trading frictions as part of the explanation. Compare the publisher's abstract with the Missouri State repository record. The finding is sample-specific, not a rule for every asset.
The boundary is part of the model
Choose timestamps that match the market you are studying. For US cash equities, one possible convention is prior regular-session close to today's regular-session open, then open to close. For instruments trading across time zones, document which venue's session defines each interval and how overlapping price discovery is handled.
| Component | Example definition | Common measurement trap |
|---|---|---|
| Overnight | Previous regular close to next regular open | Treating a gap as a fillable return without modeling the execution price |
| Intraday | Regular open to regular close | Ignoring spreads, auction mechanics or the exact timestamp of an assumed fill |
| Close to close | Previous close to current close | Comparing it with components built from mismatched sessions or adjusted prices |
The arithmetic is useful only when each leg uses compatible prices. Dividends, splits, stale prints, opening auctions and timezone changes can alter the series. A measured reversal around a session boundary may reflect the chosen prices and execution assumptions as much as an enduring market pattern.
What the evidence cannot settle
The published result is conditional on its stocks, period, session definitions and execution assumptions. It does not establish that the same pattern persists in another universe or that it can be captured after costs. A pattern may weaken when the universe, period, open-price definition or market structure changes.
The split is descriptive before it is predictive. Seeing one component dominate historically is not evidence that a trader could have known the component in advance, entered at the recorded price, or survived costs and capacity limits. Treat any proposed explanation as a hypothesis to test, not as a cause established by a chart.
How you'd actually test it
Write the measurement rule before looking at the result. For each instrument, record the exchange calendar, timezone, session boundaries, adjusted or unadjusted price choice, and the first executable quote or auction price used for a hypothetical entry. Keep overnight and intraday legs separate, then reconcile their compounded return with close-to-close data.
Next, test whether the result persists across non-overlapping periods and a universe defined point-in-time. Include spread, commission, slippage, dividends or financing where relevant. If you form a rule from one interval and tune its thresholds, reserve untouched data for evaluation and count every variant tried. The site's opening-gap explainer covers a related but narrower question: how to define and test the gap itself.
A decomposition can also be a diagnostic: if a strategy's reported edge comes almost entirely from a boundary price you could not have traded, the backtest is measuring the data convention as much as the method. realbacktesting is built around verifiable cTrader research; the same principle applies here—make the timing and fill assumptions inspectable.
Takeaway
Split the return to understand its timing, then test the executable rule separately. A session boundary is a measurement choice, not a market promise.