Quick Answer: We measured the high-to-low range of each trading session - Tokyo, London, New York and the London/New York overlap - for 28 forex pairs across each pair’s last 250 qualifying sessions of Dukascopy hourly bid candles (mid-October 2025 to September 30, 2026). London is the widest session for 26 of the 28 pairs; Tokyo is widest only for AUD/NZD (39.2 pips) and NZD/JPY (51.0 pips). On EUR/USD the London window averages 46.0 pips, New York 43.5 and Tokyo 31.5 against a full-day range of 60.4 pips. The gap between widest and quietest session runs from 1.08x on NZD/CAD to 1.70x on USD/CAD. A wider session is a longer measured move, not a better one to trade.
| Pair | Full day | Tokyo | London | New York | Overlap | Widest session | Widest ÷ quietest |
|---|---|---|---|---|---|---|---|
| GBP/JPY | 134.8 | 88.5 | 98.5 | 81.6 | 70.2 | London | 1.21x |
| GBP/NZD | 133.8 | 85.5 | 92.8 | 81.8 | 68.8 | London | 1.13x |
| CHF/JPY | 127.5 | 80.5 | 91.2 | 76.1 | 65.1 | London | 1.20x |
| GBP/AUD | 114.4 | 69.9 | 83.1 | 73.0 | 61.9 | London | 1.19x |
| EUR/NZD | 114.3 | 75.0 | 76.8 | 70.0 | 58.7 | London | 1.10x |
| EUR/JPY | 105.3 | 71.6 | 74.4 | 60.8 | 52.4 | London | 1.22x |
| USD/JPY | 103.7 | 66.6 | 71.2 | 63.8 | 52.4 | London | 1.12x |
| EUR/AUD | 98.3 | 62.0 | 69.6 | 63.1 | 54.1 | London | 1.12x |
| GBP/CAD | 94.1 | 45.7 | 74.3 | 66.4 | 54.5 | London | 1.63x |
| AUD/JPY | 91.3 | 61.6 | 62.3 | 55.6 | 47.2 | London | 1.12x |
| GBP/USD | 80.5 | 41.8 | 62.2 | 57.6 | 46.9 | London | 1.49x |
| NZD/JPY | 76.8 | 51.0 | 49.8 | 45.2 | 37.3 | Tokyo | 1.13x |
| CAD/JPY | 75.3 | 47.1 | 53.8 | 47.5 | 40.3 | London | 1.14x |
| EUR/CAD | 68.8 | 33.2 | 54.0 | 50.8 | 41.1 | London | 1.63x |
| EUR/USD | 60.4 | 31.5 | 46.0 | 43.5 | 34.9 | London | 1.46x |
| AUD/CAD | 59.0 | 35.6 | 41.5 | 39.1 | 32.3 | London | 1.17x |
| USD/CAD | 57.8 | 27.0 | 45.8 | 45.2 | 37.6 | London | 1.70x |
| AUD/NZD | 55.9 | 39.2 | 32.5 | 28.2 | 23.2 | Tokyo | 1.39x |
| GBP/CHF | 54.1 | 27.7 | 42.7 | 35.4 | 29.9 | London | 1.54x |
| AUD/USD | 54.0 | 32.4 | 38.5 | 37.4 | 30.5 | London | 1.19x |
| NZD/CAD | 54.0 | 32.3 | 35.0 | 33.8 | 26.5 | London | 1.08x |
| USD/CHF | 53.4 | 27.1 | 41.0 | 38.1 | 31.0 | London | 1.51x |
| NZD/USD | 48.8 | 29.2 | 33.0 | 32.2 | 25.6 | London | 1.13x |
| AUD/CHF | 40.6 | 23.4 | 27.9 | 24.5 | 20.7 | London | 1.19x |
| NZD/CHF | 35.1 | 18.9 | 21.1 | 19.4 | 15.5 | London | 1.12x |
| EUR/CHF | 34.2 | 18.1 | 28.0 | 23.1 | 19.7 | London | 1.55x |
| CAD/CHF | 34.2 | 15.8 | 24.8 | 23.2 | 18.6 | London | 1.57x |
| EUR/GBP | 29.6 | 15.6 | 23.9 | 20.7 | 17.6 | London | 1.53x |
As of September 30, 2026. Source: Dukascopy hourly bid candles. Figures in pips over each pair's last 250 qualifying sessions. Each session figure is the high-to-low range of that session's own UTC window on the day, not a sum of hourly ranges, so sessions that overlap in time share part of their range and the columns do not add up to the full day. Windows: Tokyo 00:00-08:59 UTC all year; London 07:00-15:59 UTC Apr–Sep and 08:00-16:59 Oct–Mar; New York 12:00-20:59 UTC Apr–Sep and 13:00-21:59 Oct–Mar. The overlap is the intersection of London and New York. The seasonal switch is by calendar month, which approximates the US and UK/EU clock changes to the nearest month. Hover a cell for the median and the session's average share of the day's range. Past ranges describe what has already happened and are not a forecast of the next session.
Setting an EA’s trading hours off this table? Our free GridMaster EA takes its spacing in pips (
GridSizePips), and the session columns above say how far a pair typically travels inside each window. Download all 7 free MT4 & MT5 EAs →
How This Table Is Computed
This is the third view of the same Dukascopy hourly bid candle cache behind the average daily range table and the hourly volatility table. The daily table reports one high-low per day; the hourly table reports the mean range of each individual UTC hour; this page sits between them and reports the high-low of each session window on each day.
A day only contributes if it passes the same filter the other two pages use: a weekday carrying at least 12 hourly bars and a non-zero daily range. For every qualifying day, the session figure is the highest high minus the lowest low across the bars whose UTC hour falls inside that session’s window, divided by the pip size (0.01 on JPY-quoted pairs, 0.0001 elsewhere). The table shows the mean over the pair’s last 250 such days; hovering a cell shows the median and the number of days behind it.
The windows are stated in UTC and switch by calendar month to track daylight saving in London and New York: London is 07:00-15:59 UTC from April to September and 08:00-16:59 from October to March, New York 12:00-20:59 and 13:00-21:59, and Tokyo 00:00-08:59 all year. That is the convention the forex trading sessions guide uses, and the month-level switch is an approximation: the US and the UK/EU change clocks on different dates, so the first weeks of each transition sit one UTC hour off for one of the two centres.
Because the sessions overlap in time, the columns are not additive. A London range and a New York range share the overlap hours, so adding them double-counts; the “share of the day” figures in the hover text are each measured against the full day independently and will sum to well over 100%.
London Is Widest for 26 of 28 Pairs
Sorted by the widest session, the table is lopsided: London takes the title for 26 pairs, and it does so on crosses with no European currency in them as readily as on EUR/USD. GBP/CAD’s London window averages 74.3 pips against 45.7 in Tokyo; USD/CAD’s London window, 45.8 pips, is almost identical to its New York window at 45.2 - Canada’s scheduled releases land at 08:30 Eastern, which is inside both windows on this clock.
The two exceptions are both Asia-Pacific crosses. AUD/NZD’s Tokyo window averages 39.2 pips against 32.5 in London and 28.2 in New York, and NZD/JPY’s Tokyo window averages 51.0 pips against 49.8 in London. Both currencies’ central banks and data releases land inside Tokyo hours, and the margin on NZD/JPY is slim enough that a different 250-day window could flip it. The other JPY crosses lean the same way without crossing over: AUD/JPY’s Tokyo window (61.6 pips) is within a pip of its London window (62.3), and GBP/JPY’s Tokyo window, at 88.5 pips, is wider than every non-JPY pair’s London window in the table except GBP/NZD’s.
The quietest session is Tokyo for 19 pairs and New York for the remaining 9. All nine have a JPY or NZD leg - six of the seven JPY pairs (CAD/JPY is the exception), plus GBP/NZD, EUR/NZD and AUD/NZD - the pairs whose own news flow has finished by the time New York opens.
How Lopsided Each Pair’s Day Is
The last column divides the widest session by the quietest. It runs from 1.08x on NZD/CAD and 1.10x on EUR/NZD - pairs whose range is spread almost evenly across the day - to 1.63x on EUR/CAD and GBP/CAD and 1.70x on USD/CAD, where the London and New York windows carry the day and Tokyo is a fraction of it. USD/CAD’s Tokyo window averages 27.0 pips, under half of its 57.8-pip day.
For the majors the ratio is middling: EUR/USD 1.46x, GBP/USD 1.49x, USD/JPY 1.12x. USD/JPY’s flat profile is the Tokyo effect again - its Tokyo window (66.6 pips) is wider than EUR/USD’s London window (46.0).
Averaged across all 28 pairs, the London window’s range is 74% of the full day’s range, New York’s 67%, Tokyo’s 61%, and the four-hour overlap’s 57%. The overlap reading lower than the two sessions that contain it is not a contradiction of the hourly table, where the overlap hours are individually the widest: a high-low grows with the length of the window, so a 4-hour range sits inside a 9-hour range that contains it.
What This Means for EA Hour Filters
Most EAs accept a trading-hours window, usually in broker server time rather than UTC, and the common mistake is to pick that window from a general “London is busiest” rule rather than from the pair being traded. This table says the rule is right 26 times out of 28 - and that the two times it is wrong, AUD/NZD and NZD/JPY, are exactly the pairs where a London-only filter would skip the widest session and trade the second-widest.
For a grid or mean-reversion EA, the more useful column is often the quietest session, because a grid spaced for the full-day range will fill fewer levels inside a window that only travels half of it. On USD/CAD, a step sized to its 57.8-pip day will fill roughly half as often in Tokyo (27.0 pips) as in London (45.8). The grid trading strategy guide covers how to turn a measured range into a spacing figure; this page only supplies the range.
Two cautions. First, server time is not UTC on most brokers - XM runs GMT+2 in winter and GMT+3 in summer - so a window read off this table has to be shifted before it is typed into an EA; the trading sessions guide covers the conversion. Second, the figures are means over 250 days that include every scheduled release in the period; a session’s typical range tells you nothing about any single day’s.
Daylight Saving and What the Month-Level Switch Misses
The London and New York windows move by one UTC hour between northern summer and winter, and this page switches them by calendar month: April-September uses the summer windows, October-March the winter ones. The real clock changes fall on the last Sunday of March and October for the UK and EU, and the second Sunday of March and first Sunday of November for the US, so for roughly one to three weeks around each change one centre’s window is an hour off in this data. The effect on a 250-session mean is small; it is stated so the figures are not read as more precise than they are.
Tokyo has no daylight saving, so its window is fixed all year and its figures carry no such approximation.
How This Differs From the Hourly and Daily Tables
The average daily range table answers “how far does this pair move in a day”; the hourly volatility table answers “which hours carry that movement”; this page answers “how far does it move inside each session”. All three run off the same cached candles with the same day filter, so a pair’s full-day figure here matches its ADR250 on the daily table to the decimal, and the sessions here are the same days the hourly table slices.
Refresh Schedule
The data file behind the table is rebuilt from Dukascopy every month, a few days after the month closes, together with the daily and hourly tables, and the as-of date under the table changes when it does. Nothing here is a forecast; most retail traders lose money trading forex regardless of which session they trade.