During the London–New York overlap, European and American desks are active together. In a hypothetical jobs-release example, EUR/USD moves rapidly and reverses; that can coincide with wider spreads and slippage rather than the day's tightest spread. The overlap is useful context, but neither a release nor a clock window guarantees favourable execution.

What the London–New York overlap is and when it runs

For an illustrative desk-day convention of 08:00-17:00 in London and 08:00-17:00 in New York, the shared interval is 13:00-17:00 London time in most weeks and 12:00-17:00 during US-UK daylight-saving mismatch periods. Convert that dated interval to your own zone. FX is decentralised, so these are comparison conventions rather than exchange opening bells.

Use named zones and the event’s date. For 2026, New York changes clocks on 8 March and 1 November; the UK and Poland on 29 March and 25 October. Under the stated 08:00–17:00 convention, the Warsaw overlap is 14:00–18:00 on most weekdays and 13:00–18:00 on weekdays 9–27 March and 26–30 October. An assumed 08:30 New York event converts to 13:30 Warsaw on 12 March and 14:30 on 2 April 2026; this is a clock example, not a claim that a report is published on either date. Provider hours, holidays and contract breaks are separate checks.

What sales-desk geography can and cannot show

As historical context, the BIS April 2022 survey reported roughly 38% of turnover at UK sales desks and 19% at US desks, with Singapore about 9%, Hong Kong SAR 7% and Japan 4%. These are geographical shares on the survey’s net-gross basis, attributed to sales-desk location rather than a particular hour or execution venue. They neither measure the overlap nor establish the spread on your account. They are 2022 figures, not current market shares.

The BIS's geographical turnover shares describe where sales desks report activity over a survey period. They do not measure the percentage traded during a London–New York overlap or prove that every order book is deepest then. More simultaneous participants can support liquidity, while news and order imbalances can disrupt it. See the sessions guide.

Sales-desk geography · historical BIS April 2022
Illustrative overlap · Europe/Warsawnormally 14:00–18:00; DST mismatch 13:00–18:00
United Kingdom share of turnoverapprox. 38%
United States shareapprox. 19%
Spread to measure on your accountEUR/USD; varies by provider and event
Sum of rounded geographical shares38% + 19% = 57%; not the overlap’s share

Compare your own dated quotes and fills instead of assuming every overlap has the lowest cost. In two hypothetical equal-interval spread samples, A is 0.8, 0.9, 1.0, 1.1, 1.2 pips and B is 0.4, 0.5, 0.5, 0.6, 8.0. A has mean and median 1.0; B has median 0.5 but mean 2.0. Dropping B’s last observation would reduce its mean to 0.5 and conceal the spike. Five observations do not estimate typical session quality or a future maximum. Keep sampling rules, account, instrument and provider consistent, including event days. The London session supplies the preceding context.

Which pairs to trade in this window

Select an instrument from its specification and measured conditions. A dollar leg does not establish a “clean” move, the tightest spread or the most decisive trend during every overlap. Compare quoted spread, available size and actual fills separately; a volatile price is not evidence of inexpensive execution.

US and Canadian announcements can coincide, but verify each publisher’s calendar and the provider’s terms for the instrument. Do not infer suitability from a major/exotic label or assume a currency stops trading when its home desks close. Use the New York session guide to organise the American part of the day, without treating it as a pair recommendation.

Two filters do not create two separate samples

One register contains ten distinct IDs. The L and NY labels indicate whether each record's timestamp falls in two assumed windows that partly overlap.

Scroll sideways to see the full diagram

Two filters can contain the same recordsA hypothetical register has three records only in L, five in both windows and two only in NY. Filter L contains eight records and NY seven. Their sum gives fifteen assignments, but removing five duplicates leaves ten unique records.TWO FILTERS, ONE REGISTEREach of the 10 records has its own ID.Filter L: 8 recordsL1–L3W1–W5N1–N2352L onlyBoth filtersNY onlyFilter NY: 7 records8 + 7 = 15 filter assignments8 + 7 − 5 = 10 unique records

Hypothetical register: L1–L3 only in L, W1–W5 in both windows, N1–N2 only in NY. Each record appears once in the register. Filter names do not attribute transactions to a country or execution venue. The document TradingView — Sessions confirms that membership in a custom window can be determined from the record's timestamp and an explicitly defined session.

The middle five records belong to both filters. Assignments to windows and unique records answer different questions.

To examine all records from at least one window, take the union of their IDs. When comparing windows, their shared portion means that the samples are not disjoint. Do not treat the same five observations as ten independent pieces of evidence without justification. Deduplicating by a rounded timestamp is insufficient: two different records can share a timestamp, so this model uses the ID as its key.

Scheduled events and execution uncertainty

The Employment Situation report and CPI have dated BLS schedules, generally with publication at 08:30 US Eastern time. The jobs release is not invariably on the first Friday. In Europe/Warsaw, 08:30 Eastern normally converts to 14:30, and to 13:30 during the brief daylight-saving mismatch. Check the actual date and official calendar.

OANDA Europe Limited’s support page describes dynamic spreads widening around news, changes in liquidity and uncertainty, with rapid or gapped price changes possible. This named provider’s explanation is enough to reject an unconditional “overlap must be cheapest” rule; it supplies no universal pip multiplier or waiting period. Set your response to positions and pending orders before the event. Waiting does not prove that a “true direction” has emerged, and a central-bank announcement has its own calendar rather than the jobs-report timetable.

The realistic slice of the overlap for someone working a day job

Choose a monitoring period from the dated overlap and your actual availability. In a hypothetical plan, 16:15–17:45 gives 90 minutes; excluding a preselected 16:25–16:40 event window leaves 75 minutes. The exclusion is an exercise input, not a universal news buffer or instruction to trade whenever time remains.

Keep timestamps and specify what is equally weighted. In a separate hypothetical hour, spread is exactly 1 pip for 59 minutes and 7 pips for the last minute. Ten records from the first segment and ninety from the last give a record-weighted mean (10 × 1 + 90 × 7) / 100 = 6.4 pips. A time-weighted mean is (59 × 1 + 7) / 60 = 1.1 pips. Neither alone gives your strategy’s realised trading cost: actual order times, commission, financing and fills matter. Alerts notify; pending orders can create exposure and need separate controls. See best hours to trade.

What to do tomorrow

  1. Separate session convention and instrument hours. Define the two local desk windows, convert them with the date and named zones, then check your actual provider’s holidays and contract breaks. A broker timetable does not define the world’s overlap.
  2. Compare costs at three times. Record spread, commission and actual slippage for the same instrument and size in several dated windows. Use the observations to compare conditions; do not predetermine that the afternoon must be cheapest.
  3. Record actual release times. Take the jobs and CPI dates from BLS, store the US Eastern time and convert it for your local zone on each date. Most weeks that is 14:30 in Europe/Warsaw; during the DST mismatch it is 13:30.
  4. Match the window to your availability. Convert the dated overlap to your own zone, then compare it with the hours you can monitor. Select pairs and exposure limits from your plan and actual provider conditions rather than assuming a European after-work slot.