Early access is open. Free sample datasets are live now. Download a sample

Timestamps, Time Zones and Trading Sessions

Details

Docs navigation

Time is where most data bugs hide: bars shifted an hour after daylight saving, or a futures session that starts the evening before. This page sets out the rules we follow for market data timestamps in UTC, how daily dates are assigned, and which sessions each asset class has. Read it before you merge our data with anything else.

Early access. The rules below apply to the planned v1 API and to every file in our free sample library.

Details

The rules

1. Every timestamp is UTC, ISO 8601

2026-11-18T14:31:00Z                  bar start, 1-minute stock bar
2026-11-18T14:30:01.204518337Z        stock trade, nanoseconds
2026-11-19T14:32:07.418Z              FX tick, milliseconds

The trailing Z means UTC. We never send local times in a timestamp field.

2. Bars are stamped at the start

A 1-minute bar stamped 14:31:00Z covers 14:31:00.000 up to but not including 14:32:00.000. Some vendors stamp at the end of the bar, so their 14:32 is our 14:31. If you merge sources, check this first.

When nothing trades in an interval, we do not emit a bar. Thinly traded symbols will have gaps. Reindex in pandas if you need a full grid.

3. Daily data uses a date, not a timestamp

Daily rows carry date as YYYY-MM-DD in the exchange's own calendar. For US stocks, 2026-11-18 is the New York session of that day. There is no time part, so it cannot drift with time zones.

4. Each asset class defines its own day

Asset class Session times What date means
US stocks Pre-market 4:00 to 9:30, regular 9:30 to 16:00, after-hours 16:00 to 20:00, New York time New York calendar day
US options 9:30 to 16:00 New York for most; some index options trade longer New York calendar day
Futures (CME Globex) Most products: Sunday to Friday, 17:00 to 16:00 Chicago time, one-hour daily break session_date: the trading day the bar belongs to, which opens the evening before
Forex Sunday evening to Friday evening, around the clock Day closes at 17:00 New York, shown in day_close_time
Crypto 24/7, no close UTC calendar day, 00:00 to 24:00 UTC
Indices Follow the underlying exchange Exchange calendar day

A Globex bar at 2026-11-17T23:00:00Z (Monday evening in Chicago) has session_date 2026-11-18. Group by session_date, not by the UTC date of timestamp, when you build daily futures bars.

5. Intraday stock bars carry a session field

Values are pre, regular, post and overnight. Filter on it instead of on clock times:

Example (Python):

regular = bars[bars["session"] == "regular"]

6. Convert to local time only for display

Example (Python):

import pandas as pd

bars["timestamp"] = pd.to_datetime(bars["timestamp"], utc=True)
bars["ny_time"] = bars["timestamp"].dt.tz_convert("America/New_York")

Do your joins and resampling in UTC. Convert at the end.

Details

Daylight saving, worked through

The US moves clocks on different weekends from Europe, and UTC never moves. So the 9:30 New York open falls at:

Period New York offset Open in UTC
November to March (standard time) UTC minus 5 14:30Z
March to November (daylight time) UTC minus 4 13:30Z

For two or three weeks each spring and autumn, the gap between New York and London is four hours instead of five. FX and futures strategies keyed to "London open" will shift against US data in those weeks. Pandas handles this if you convert with a named zone (America/New_York, Europe/London) rather than a fixed offset.

Watch out

Notes and gotchas

  • start and end parameters accept either a date (2026-11-18) or a UTC timestamp. A date alone means 00:00 UTC on that day, which is 19:00 New York the evening before in winter. For one stock session, pass UTC times.
  • Early closes (for example the day after US Thanksgiving) end at 13:00 New York. The market calendar lists them with status: "early_close".
  • Exchange holidays return empty results, not errors.
  • Tick data can have two times: timestamp (SIP or exchange time) and participant_timestamp or received_timestamp. Sort on timestamp and then on sequence_number for stocks.
  • Our free 1-minute BTC-USDT sample follows these rules: 500 bars, each exactly 60 seconds apart.

Keep reading

Next steps

Details

Was this page useful?

[Yes] [No]

Have a time zone case we did not cover? Tell us on the custom data form.