The best time to trade forex, measured

Every answer you will find to this says the London–New York overlap, and every answer is right — for a reason most of them state wrongly. That window moves 2.41× a typical hour, which is a fact about how far price travels and about what it costs you to trade. It is not a fact about which way price goes, and the difference is the whole of this article. From 23 years of hourly bars across 19 markets.

The short answer

If you are choosing hours, choose them for movement and cost, which are measurable and stable, rather than for direction, which barely survives its own test. In UTC that means roughly 12:00–16:00, when London and New York are both open. The rest of this page is the evidence, the per-market version, and the hours to stay away from.

Every hour of the day, sized

Each hour below is the median across our markets of how far that hour moves, as a multiple of that market's own typical hour — so 47 points of an index and 10 pips of a currency pair are on the same scale. Monday to Friday, in UTC.

The peak is 14:00 UTC at 1.72×; the quietest is 04:00 at 0.57×. That is a spread of roughly 3.0 to 1 between the busiest and quietest hour of an ordinary weekday — which is why the question is worth asking at all.

Why the overlap is a cost argument, not a direction one

Here is the part almost nobody publishes. The same 23 years, split by session, asking two different questions of each one:

The half that survives a holdout, and the half that does not

Any of this is only worth reading if it holds on data it was not chosen on. We ran that test twice, on the two axes separately.

• Timing survives. Rank the hours using only pre-2020 candles, then score that ranking on 2020 onward: the median market comes back at 0.925, and the ten hours picked before 2020 were still worth 1.82× a typical hour afterwards against 0.99× for every other hour. ⚠ It even holds inside the busy band alone (0.859), where every hour is already busy — which is the control that stops this being "London is busy" restated.
• Direction barely does. Windows picked on pre-2020 candles by our own significance bar were worth +0.56 percentage points in ordinary hours afterwards (z = 1.62), which is less than a round trip. The control — 280,454 holdout candles from windows that were not picked — reads 0.01, exactly as it should.

So: the timetable is a real, reusable object. The arrow is not.

Every market's busiest window

Ranked by how far that window moves against the market's own ordinary hour, with the round trip beside it — because a window that moves three times its usual amount is still not worth trading if the spread eats it. All UTC.

Each market name links to its own article.

The hours to stay away from

The daily rollover — 21:00–22:00 UTC — is where this data produces its most impressive-looking numbers and its least usable ones. Liquidity is thin, the spread is at its widest, and our own validation flagged those hours as artefacts of the session boundary rather than as edges.

What "the best time" cannot tell you

• It is a statement about when, never about where. Knowing the 13:00 hour typically moves 12 pips tells you how to size a stop. It tells you nothing about the level, and nothing about the direction.
• A busy hour is a risky hour. The same volatility that makes a window worth trading is what takes you out of it. These are long-run medians, not forecasts for this afternoon.
• Your clock is not our clock. Everything here is UTC because that is the only frame in which all our markets can be compared. Convert it once, carefully — a weekday and an hour only mean something together, and crossing midnight moves the weekday too.
• Check it against your own spread. Our costs are a modern typical, not your broker's. The dashboard and the screener both let you set your own and watch the list shrink.

The method behind every figure above is on the evidence page, and every claim this industry makes about timing — including the ones that failed — is on the reports index.

More on when a market actually moves

Every claim we have tested, including the ones that held, is on the reports index.