Does New York follow London's direction in currency crosses?
- Markets
- Period
- –
- Period length
- 2005 – 2026
- Data
- 1-hour candles built from one-minute data
- Sample
- 118,402 market-days from 21 markets (about 5,640.0 trading days each)
- Source
- OANDA CFD one-minute mid prices · New York trading days
A session's direction is simple: did it close above or below its open? We asked whether the next session tends to go the same way, for every trading day of 21 markets from 2005 to September 30, 2026: EUR/GBP, EUR/JPY, EUR/CHF, EUR/AUD, EUR/CAD, EUR/NZD and 15 more (all listed in the results below).
Forex session times used here (New York time)
Every trading day runs from 17:00 to 17:00 New York time and is cut into three sessions that do not overlap, so each hour belongs to exactly one of them:
- Asia: 17:00 to 02:00 (9 hours, starting the evening before)
- London: 02:00 to 08:00 (6 hours, before New York opens)
- New York: 08:00 to 17:00 (9 hours, including the London afternoon)
The sessions are not the same length. If time were all that mattered, Asia and New York would each get 37.5% of anything and London 25%. Every result below should be read against those numbers.
For this study the session lengths do not matter: it only compares the direction of one session with the next.
New York after London
Taken together, New York closed in the same direction as London on 49.9% of days. Market by market, the lowest was NZD/CAD at 47.8% and the highest AUD/CHF at 51.1%. 20 of the 21 markets are within two points of 50%.
London after Asia
London followed Asia's direction on 49.1% of days, a coin flip. Markets in the same group move together, so the totals overstate how sure we can be.
Year by year
Over the 21 full years in which all 21 markets traded, the yearly share for London to New York stayed between 47.7% (2008) and 52.4% (2005), on both sides of 50%, so there is no steady lean in either direction. The markets of a group move together, so swings of a few points in a single year are ordinary noise. The direction of one session tells you essentially nothing about the direction of the next.
Why publish a result that found nothing?
"London sets the direction for the day" is a popular idea, and this is what it looks like when counted on years of data. A clear "no" is useful: it saves time, and any rule built on session direction has to beat these numbers, after costs, before it is worth anything.
Results by market
All years together, one row per market; the last row adds the markets together. Every value is a percentage of the days counted. Pick a market's name to see it year by year.
Crosses
21 markets
| Market | Days | London → New York | Asia → London |
|---|---|---|---|
| EUR/GBP | 5639 | 49.8 | 49.1 |
| EUR/JPY | 5638 | 49.8 | 49.8 |
| EUR/CHF | 5639 | 50.2 | 48.0 |
| EUR/AUD | 5639 | 50.8 | 48.3 |
| EUR/CAD | 5639 | 50.2 | 49.0 |
| EUR/NZD | 5636 | 49.8 | 49.5 |
| GBP/JPY | 5639 | 49.0 | 50.0 |
| GBP/CHF | 5639 | 51.0 | 50.2 |
| GBP/AUD | 5639 | 49.9 | 47.9 |
| GBP/CAD | 5636 | 49.9 | 47.5 |
| GBP/NZD | 5639 | 49.3 | 49.4 |
| AUD/JPY | 5639 | 50.1 | 50.0 |
| AUD/CHF | 5639 | 51.1 | 49.2 |
| AUD/CAD | 5636 | 49.4 | 47.7 |
| AUD/NZD | 5638 | 48.8 | 49.1 |
| CAD/JPY | 5639 | 50.4 | 49.6 |
| CAD/CHF | 5639 | 50.3 | 47.6 |
| CHF/JPY | 5639 | 49.3 | 49.3 |
| NZD/JPY | 5636 | 49.8 | 50.3 |
| NZD/CHF | 5639 | 50.5 | 50.1 |
| NZD/CAD | 5636 | 47.8 | 50.1 |
| All crosses | 118402 | 49.9 | 49.1 |
Year by year
EUR/JPY: 5638 complete trading days, 2005-01-03 to 2026-09-30.
- The line in each bar marks 50%, a coin flip.
- Bars, top to bottom: new york closes in the same direction as london; london closes in the same direction as asia.
| Year | Days | London → New York | Asia → London | Chart |
|---|---|---|---|---|
| 2005 | 260 | 48.6 | 46.3 | |
| 2006 | 259 | 47.5 | 54.9 | |
| 2007 | 259 | 51.7 | 48.6 | |
| 2008 | 260 | 45.4 | 50.0 | |
| 2009 | 259 | 53.5 | 49.2 | |
| 2010 | 260 | 53.1 | 46.9 | |
| 2011 | 260 | 47.3 | 48.3 | |
| 2012 | 259 | 47.9 | 52.3 | |
| 2013 | 259 | 50.0 | 51.9 | |
| 2014 | 259 | 47.3 | 50.4 | |
| 2015 | 259 | 48.1 | 46.5 | |
| 2016 | 259 | 54.1 | 47.5 | |
| 2017 | 258 | 51.2 | 51.9 | |
| 2018 | 259 | 46.7 | 55.4 | |
| 2019 | 259 | 49.2 | 48.6 | |
| 2020 | 260 | 48.3 | 45.3 | |
| 2021 | 260 | 50.8 | 53.8 | |
| 2022 | 258 | 50.0 | 49.4 | |
| 2023 | 259 | 52.5 | 50.2 | |
| 2024 | 260 | 51.0 | 48.6 | |
| 2025 | 259 | 49.6 | 51.4 | |
| 2026* | 194 | 51.8 | 48.5 | |
| All years | 5638 | 49.8 | 49.8 |
Every value is a percentage of the days counted. * 2026 runs until 2026-09-30.
How this was measured
- Data: OANDA CFD one-minute mid prices (halfway between bid and ask). They are grouped into 1-hour candles in New York time, so US daylight saving time is followed automatically, including the older rules before 2007.
- Trading day: 17:00 to 17:00 New York time, Monday to Friday. Asia 17:00–02:00, London 02:00–08:00, New York 08:00–17:00.
- Which days count: a day needs at least one traded hour in every session and at least 75% of the usual number of trading hours for that market in that year. Holidays and half days drop out this way.
- Which years count: a market's year is used only if it has at least 100 complete days. Years where a market did not trade in all three sessions are therefore left out for that market.
- Ties: when two sessions reach exactly the same price, the earlier session gets it.
- Totals: "together" simply adds up the market-days of the 21 markets. Markets in the same group move together (the crosses share currencies, the US indices move as one), so 118,402 market-days are not 118,402 independent observations.
- Year by year in the text: only full calendar years in which every one of the 21 markets has data are used, so the group is the same in every year. The current, unfinished year is not part of that range (it is in the tables, marked with *).
- Period: 2005 (or the first year with full data for a market) to September 30, 2026.
- Direction: a session is up if its close is above its open, down if below. Sessions that close exactly at their open are left out of that comparison.
Historical statistics for research and education, not trading advice. Past market behaviour does not guarantee future results.