Trading costs can vary by time, instrument, account and order size. A quote that looks inexpensive at one time need not remain available later. Compare spread records and execution costs before selecting a trading window; the clock alone does not establish whether a trade has an edge.

What forex liquidity actually means

Liquidity concerns the ability to transact at available prices and sizes. A displayed Bid–Ask spread is one measure, but it does not guarantee that an order of a given size will fill at that price. Check quote depth where available and compare actual fills with the quote observed when the order was submitted.

BIS describes spot and most FX derivatives as decentralised OTC markets with multiple dealers and venues. Local order books and request-for-quote systems exist, but no one book contains all FX liquidity. A country’s survey turnover share does not identify your account’s hourly spread or available size.

Tokyo hours: define the observation window

For comparison, define Tokyo as 09:00-18:00 Asia/Tokyo, or 00:00-09:00 UTC. Convert this window for your local zone and date. Other Asian centres have their own hours; measure your instrument and account rather than assuming non-yen pairs cannot be traded then.

Use hypothetical spreads to understand the cost calculation, not as broker averages. If a quote changes from 0.3 to 1.2 pips, the difference is 0.9 pip. Convert it using the position-specific pip value, then account separately for commission, slippage and financing.

London hours: measure depth for your order size

Define London here as 08:00-17:00 Europe/London, an illustrative desk window. Convert its boundaries for the date and your local zone. BIS country-level desk statistics do not prove an individual account’s spread, depth or profitability during that window.

Sterling pairs can respond to UK releases and order flow during London hours. Record their observed range, spreads and fills over several dates. Neither the time of day nor a large range establishes a profitable trade direction.

The London–New York overlap: a dated intersection

Using London 08:00–17:00 and New York 08:00–17:00 local windows, the overlap is 13:00–17:00 London time for most of the year and 12:00–17:00 during US–UK daylight-saving mismatch periods. In 2026, the UK changes clocks on 29 March and 25 October; New York changes on 8 March and 1 November. On 16 March and 28 October, the overlap is 12:00–17:00 London time; on 15 January and 15 July it is 13:00–17:00. These are conversions of chosen windows, not measured liquidity results.

An overlap identifies simultaneous local working windows. It does not guarantee a specific spread or absence of slippage. Compare quotes and fills on the same instrument, account and size; evaluate Tokyo–London handover conditions separately.

Financing cutoffs and trading breaks

Read trading availability and financing separately. For OANDA Corporation’s US FX service, most pairs trade from Sunday 17:05 to Friday 16:59 New York time, with a daily 16:59–17:05 break; TRY and NZD have specific exceptions. Its overnight financing assessment uses 17:00 New York time. Another account or instrument can use a different schedule; server midnight is not a substitute for the stated cutoff.

Quotes and available sizes may change around a provider cutoff. Spread and slippage are separate: a wider quote is visible before entry, while a fill can differ from the submitted or observed price. Use your own records to assess the cutoff; adverse execution can also occur at other times.

Two readings do not represent two equal time intervals

A record average describes records. To describe time spent at a spread, you need its full path and interval lengths.

Scroll sideways to see the full diagram

Two readings do not represent two equal time intervalsA fictional complete spread path is 1 pip for the first 50 minutes and 5 pips for the last 10. The 0–60 axis uses proportional intervals. Two change records average 3 pips; the time-weighted mean is 100/60, about 1.67 pips. These are not market research results.Records and minutes have different weightsOur model: the exact spread is known for the whole hour.1 pip × 50 min5 pip0 min50 min60 minTwo records(1 + 5) / 2 = 3 pipsElapsed time(50 + 50) / 60 ≈ 1.67Same history. Different question for the data.
Explicit model with two constant intervals; the diagram does not estimate spreads in missing data.

Assumptions: a fictional hour 00:00–01:00 on one clock. The full path is known in advance: exactly 1 pip for 50 minutes, then exactly 5 pips for 10 minutes. The export records only two changes: at the start and at minute 50. There are no gaps, missing data or executed orders.

The mean of the two recorded values is (1 + 5) / 2 = 3 pips. The time-weighted mean is (1 × 50 + 5 × 10) / 60 = about 1.67 pips. The first calculation gives each record equal weight; the second gives each minute equal weight. Merely duplicating unchanged records would alter the first result while leaving the spread path unchanged.

Do not infer a long constant interval from two distant timestamps alone: constancy is an explicit assumption here. OANDA pricing documentation separates quote creation time, Bid, Ask and tradeability. For your own measurement, check completeness and the recording rule. Neither mean above gives the cost of an actual trade.

A hypothetical EUR/USD cost comparison

The following is our hypothetical EUR/USD model, not a measured session average: buy EUR 100,000 and immediately sell it with unchanged quotes. A pip is 0.0001 USD per EUR, worth USD 10. Both quotes have midpoint 1.10000. Assume full fills at the displayed Ask and Bid, USD 6 total commission, and no slippage, financing or account-currency conversion.

Hypothetical spread comparison

Quote A is Bid 1.0999875 / Ask 1.1000125: the 0.25-pip spread gives (Bid − Ask) × 100,000 = −USD 2.50, or −USD 8.50 after commission. Quote B is Bid 1.09990 / Ask 1.10010: its 2-pip spread gives −USD 20, or −USD 26 after commission. These precision and fill assumptions serve only the model; no broker’s minimum tick or quote is asserted.

The net-cost difference is USD 17.50. The larger spread is eight times the smaller one, but the total costs are not in that ratio because the same commission applies to both. Do not deduct the spread again from an Ask-to-Bid result. Scaling USD 17.50 to 240 identical round trips gives USD 4,200 only if every assumed spread difference repeats; it is not an observed monthly saving or proof that another hour provides the same opportunity.

What to do tomorrow

  1. Record simultaneous Bid and Ask quotes with timestamps and check whether the export is complete. A Bid-only candle cannot reveal the historical spread. Different spread observations show variation at those times; no fivefold threshold is required, and three readings do not establish a lasting session pattern.
  2. Choose candidate entry windows that fit your schedule. Compare their spreads and execution records over multiple dates before adopting a default. The London–New York overlap is a time-zone intersection, not a guarantee of the cheapest or safest entry.
  3. Mark your own provider financing cutoff and trading break, if any, using dated local conversions. Review whether your strategy and orders can tolerate the observed spread and execution conditions. Do not use a universal 23:00–00:00 CET restriction.
  4. Match instrument selection to the strategy, available quotes, relevant releases and your capacity to supervise orders. Local business hours provide context but do not impose a universal rule that a currency pair may only be traded during its home banks’ hours.

Related reading: the London session — local desk hours and dated conversions; the New York session — US announcements and provider conditions.