swapsPRO

Docs · Tools

Wall Street closes. These don't.

The New York Stock Exchange shuts at four o'clock and stays shut all weekend. Coinbase's tokenized equities on Base do not. Night Market shows what each company has done since the closing bell — and for somebody whose evening is New York's night, that is most of the hours they are awake to trade.

The clock is derived, not pasted#

A hardcoded list of one year's market holidays is right until January and then silently wrong — it would quote a close that never happened, with nothing on screen saying so. Every one of the 10 full-day closures comes from its rule instead: the third Monday in January, the Friday before Easter, the observance that shifts a fixed date off a weekend.

  • Regular session 09:30–16:00 Eastern, and the two abbreviated 13:00 sessions — the Friday after Thanksgiving and Christmas Eve when it falls on a weekday.
  • Daylight-saving transitions are computed against UTC instants, not local time.
  • The last close is found by walking backwards, so a long weekend is crossed correctly: on Labor Day the reference is Friday, and the next open is Tuesday.

What the clock cannot model

Unscheduled closures. The market has shut for a state funeral, a hurricane and a terrorist attack, and no rule predicts those. On such a day the page believes there was a session and reports a close that did not happen. The cost is one wrong reference price on a rare day; a hand-maintained list would be wrong far more often.

Where the move is measured from#

Both prices come from the same source: one reading at the instant of the last regular close, one now. The source answers with the nearest observation rather than the exact instant asked for — a 4:00pm question can come back with a 3:51pm reading — so the page prints the timestamp it actually used. The whole claim is “since the close”, and hiding that gap would make the claim slightly false.

What is left out, and why it is named#

An equity we cannot price on both sides of the bell is left out of the list entirely rather than drawn at 0.00%. On a row that already carries a dollar figure, “0.00%” reads as “we checked, it has not moved” — a claim nobody made. The ones left out are named on the page, so nine silent absences do not read as a bug.