Holidays, Half-Days And Thin Sessions
Trading calendars aren't uniform — market holidays, early-close half-days around holidays, and structurally thin sessions all distort volume, spreads and volatility in ways a naive backtest or live system can misread as a genuine signal.
Markets don't trade a uniform amount every day. Full holidays close the exchange entirely; half-days — the session before Thanksgiving and Christmas Eve in the U.S. equity market, for example — close hours early with sharply reduced volume in the shortened session; and even ordinary full sessions around a long weekend or during a low-liquidity summer week can be structurally thin, with wider spreads and choppier price moves simply because fewer participants are active, not because anything informative happened.
A system that doesn't account for this calendar structure can misfire in several ways: a backtest that treats a half-day's return the same as a full day's return will distort volatility estimates and any time-of-day-based signal; a volume-based liquidity filter run on a holiday-adjacent thin session might flag a stock as illiquid when it's actually normal, just quiet that particular day; and a live system that doesn't know tomorrow is a half-day might size or time an order assuming six and a half hours of trading when only three are actually available.
Concrete illustration. A U.S. equity execution algorithm scheduled to work an order evenly across a "normal" 6.5-hour session gets deployed on the day after Thanksgiving, a half-day that closes at 1pm. Without an explicit trading-calendar lookup telling it this, the algorithm plans its participation schedule against the wrong close time and ends up either rushing to finish or leaving size unexecuted when the market shuts three and a half hours earlier than its internal assumption expected.
Trading calendars — full holidays, early-close half-days, and structurally thin low-liquidity sessions — must be handled explicitly in both backtests and live systems; treating every trading day as an identical 6.5-hour full session silently corrupts volume, spread, and volatility measures on exactly the days that differ most from that assumption.
Further reading
- Practitioner convention on exchange trading calendars