calendar

name	calendar
description	Use whenever a strategy depends on the market calendar — market closures/holidays, trading-day counting, the Nth trading day before/after a date, or whether a date is a trading day. Use QuantConnect's built-in calendar; do not hand-roll a holiday calendar or hardcode dates.

Market calendar — use QC's built-in calendar, don't reinvent it

Don't hand-roll a trading-day loop or hardcode dates. QuantConnect exposes the real historical calendar for each security's exchange. But it does NOT hand you a clean "named holidays" list, so read this carefully — the obvious-looking APIs do something different from what their names suggest (all behaviour below is verified on the cloud).

Stepping / counting trading days — THIS is the reliable primitive

On the exchange hours object:

hours = self.securities[symbol].exchange.hours
prev = hours.get_previous_trading_day(d)   # -> datetime, trading day strictly before d
nxt  = hours.get_next_trading_day(d)        # -> datetime, trading day strictly after d
open_today = hours.is_date_open(d)          # -> bool, True iff the exchange trades on date d

These skip weekends and all closures, so no manual weekend/holiday checks are needed.

• "Last trading day before date H" → hours.get_previous_trading_day(H).
• "Nth trading day before H" → start at H and call get_previous_trading_day N times in a loop.
• Verified: before Memorial Day 2025 (Mon 5/26), one step → 5/23, five steps → 5/19; is_date_open(5/26) → False.

The session open/close TIME on a date (early-close aware)

For the actual TIME the exchange opens/closes on a date — not just whether it trades — use the exchange-hours helpers, instead of reading .early_closes or hand-rolling the session segments:

hours = self.securities[symbol].exchange.hours
close_dt = hours.get_next_market_close(self.time, False)   # next close at/after self.time; extended_hours=False
open_dt  = hours.get_next_market_open(self.time, False)

These return the REAL boundary for that date, so they already account for early-close half-days (e.g. a 13:00 close) and holidays. Use them for time-to-close logic, or to find when "the close" actually is, rather than hardcoding 16:00 or computing it from the hours segments yourself.

Full-day market closures — exchange.hours.holidays

self.securities[symbol].exchange.hours.holidays → a set of datetime (midnight) of EVERY full-day closure for that exchange. Compare on the date component: d.date() in {h.date() for h in hours.holidays}.

This is NOT a list of the named federal holidays. It is the union of ALL full-day closures, which includes:

• Good Friday (a market closure but not a federal holiday), and
• ad-hoc closures — e.g. 2025-01-09 was closed for a national day of mourning; historically also hurricanes, 9/11, etc.

So you cannot get "the nine federal holidays" by taking .holidays and dropping Good Friday — you would still pick up the ad-hoc days. Use .holidays / is_date_open to answer "is the market closed on date D," not to enumerate named holidays. Half-days are separate: .early_closes / .late_opens (dicts) — a half-day is not in .holidays.

Pitfall — get_days_by_type does NOT return holidays

self.trading_calendar.get_days_by_type(TradingDayType.PUBLIC_HOLIDAY, start, end) returns ALL non-business days (weekends + closures) — ~117 for a single year, not ~10. The TradingDay.public_holiday flag is likewise True for weekends (a Saturday comes back with public_holiday=True, weekend=True). Do not use this method, or that flag, to find holidays.

• self.trading_calendar.get_trading_days(start, end) is fine for enumerating days; identify a weekday closure as business_day == False and weekend == False (this still includes Good Friday + ad-hoc closures, same caveat as .holidays).

When a strategy targets SPECIFIC NAMED holidays

QC's calendar can tell you whether a day is closed and can count trading days, but it cannot name a closure. If the method needs particular named holidays (e.g. the nine U.S. federal holidays, excluding Good Friday), compute each holiday's NOMINAL calendar date by its rule (a fixed date like datetime(y, 7, 4), or an Nth-weekday like the 3rd Monday of January; honor start years such as Juneteenth from 2022). Then ANCHOR on that nominal date and let the calendar API map it to trading days:

• "Last trading day before holiday H" → hours.get_previous_trading_day(nominal_H). It already skips the weekend, the observed closure, and any adjacent closures, so it returns the correct day in every case.
• Do NOT hand-roll observed-date shifts (Saturday→Friday / Sunday→Monday). The generic shift is WRONG for New Year's: NYSE does not close the preceding Friday when Jan 1 is a Saturday (verified — Dec 31 2021 was a normal trading day, with is_date_open True and not in .holidays), so a hand-rolled shift mis-dates the window by a day. The nominal-anchor approach is correct for all of them, New Year included, with no special cases.

Do not try to reverse-engineer names out of .holidays.