2012issue C0234-39
Overnight versus daytime return evaluation
Split one ordered price series into overnight and daytime session windows, compare an explicit close-to-open rule with the open-to-close path, and keep the bias only if a cost-adjusted baseline, shorter clocks, and an autocorrelation test still support it.
- Define overnight return as the close-to-open change and daytime return as the open-to-close change on the same ordered price series.
- A naive overnight hold can produce thousands of trades, so restate the same close-to-open rule as a cost-adjusted baseline before treating the raw edge as tradable.
- When the overnight interval is split into post-market and premarket session windows, most of the measured return can sit in the early premarket clock rather than after the close.
- An autocorrelation test between daytime and overnight returns was negative for almost every symbol in the archive panel, so a strong regular session did not reliably continue overnight.
Two windows on the same series
A close-to-open overnight window can be defined separately from an open-to-close daytime window on the same ordered price series. Overnight return is the change from one session close to the next session open. Daytime return is the change from a session open to that same session close.
A session window is a fixed clock interval used to isolate premarket, regular-hours, or post-market returns.
What one long daily sample showed
On one long SPY daily sample, the overnight window showed a positive annualized return while the daytime window showed a negative annualized return.
In that same comparison, overnight variability and the reported Sharpe-style ratio were more favorable than the daytime window.
Restate the overnight hold after costs
A naive overnight hold produced thousands of trades, and a one-cent-per-share commission cut net profit sharply. A two-cent commission or added slippage was enough to erase the leftover profit.
A cost-adjusted baseline is the same close-to-open rule restated after a per-share commission so the raw edge can be compared with a tradable result.
Check shorter clocks inside the overnight interval
When the overnight interval was split into post-market and premarket clocks, most of the measured return sat in the early premarket window rather than after the close.
Repeat the split across a panel
Across a panel of sector and style ETFs, average overnight returns were higher than average daytime returns, with the widest gaps in utilities, technology, energy, industrials, and homebuilders.
An autocorrelation test is a signed check of whether one session window's return tends to continue or reverse in the next window. An autocorrelation test between daytime and overnight returns was negative for almost every symbol in the panel, so a strong regular session did not reliably continue into the overnight window.
Editorial reading
TradersWeek editorial view: compare an explicit close-to-open rule with the open-to-close path on the same series. Keep the overnight bias only if the cost-adjusted baseline, the shorter session windows, and the same-series autocorrelation test still support it. This reading is editorial and is not part of the archive facts.
All readings on this track · 15 readings
- 1987Evaluating weekly price cycles with autocorrelation cutoffs
- 1988Serial correlation is not an independence test
- 1988Event-triggered chi-square tests for jittered serial links
- 1988Reverse the rule before you trust the system
- 1990Chi-square evaluation of lagged breadth correlations
- 1990Two-gate out-of-sample tests for trading rules
- 1990Sign-cell correlation, chi-square checks, and lag shifts
- 1998Test serial dependence before you optimize a mechanical system
- 2003How a one-rank rotation is constructed for an offset autocorrelation check
- 2004Constructing first-order residual autocorrelation tests
- 2004A two-stage Durbin-Watson residual screen
- 2011Constructing trend labels from increment correlation atoms
- 2012Overnight versus daytime return evaluation
- 2012Constructing a multi-horizon folded-correlation reversal filter
- 2016Tuning indicators to measured market cycles