Selecting trading hours requires both a practical schedule and cost records. Compare candidate windows on the same instrument and account. A quiet chart, a low spread or a convenient hour alone does not establish that a method is profitable.

Define hours before comparing costs

Retail FX provider schedules commonly span working days, with exceptions for holidays and instrument-specific breaks. Activity and quotes vary within that schedule. Check the current provider timetable and measure your own costs rather than inferring them from which region is awake.

Define the London window as 08:00–17:00 Europe/London and New York as 08:00–17:00 America/New_York. These are comparison conventions, not a broker trading timetable. Convert dated boundaries into your own zone: UK and US daylight-saving change dates differ, and Europe/Warsaw uses CET or CEST according to season.

Which hour is cheaper depends on the measure

State the result for a measure chosen before comparison. The median and mean can rank the same windows in opposite orders.

Scroll sideways to see the full diagram

Which hour is cheaper depends on the measureEach fictional window has four equally weighted spread readings. A: 1, 1, 1 and 9 pips; B: 2, 2, 2 and 2. All eight bars share a zero-to-ten-pip scale. A has median 1 and mean 3; B has median 2 and mean 2. The one wide reading remains in the data.Same sample, two rankingsShared 0–10 pip scale. Each record has equal weight.Window AWindow B1112139421222324Median: 1 pipMean: 3 pipsMedian: 2 pipsMean: 2 pipsOne wide observation stays in the data.
Four fictional, equally weighted readings per window. Bars display all data without removing extremes.

Assumptions: fictional windows A and B use the same pair, account, quote size and sampling frequency. Each has four readings: A = 1, 1, 1, 9 pips; B = 2, 2, 2, 2 pips. These are quote readings, not executed trades or measurements of real sessions.

The median of A is (1 + 1) / 2 = 1 pip, below B's 2. The mean of A is (1 + 1 + 1 + 9) / 4 = 3 pips, above B's 2. A has the lower middle reading; B has the lower mean. NIST distinguishes mean and median; individual extreme values affect these measures differently.

Record both measures and show extreme readings and gaps. Do not remove 9 pips merely because it changes the ranking; a data-error correction requires an independent reason and an audit trail. Four observations cannot establish lasting conditions. Neither the median nor the mean spread determines your strategy's profitability.

A candidate window: the London–New York overlap

With those conventions, the overlap is 13:00–17:00 Europe/London for most of the year and 12:00–17:00 during US–UK clock-change mismatches. In 2026, New York changes on 8 March and 1 November; London changes on 29 March and 25 October. On weekday dates 9–27 March and 26–30 October, the chosen intersection lasts five hours rather than four. These calculations describe defined desk windows, not measured liquidity or an individual contract’s trading availability.

Price movement and spread are different measurements. Our hypothetical EUR/USD comparison buys and sells EUR 20,000 on a USD account. Midpoint rises from 1.10000 to 1.10060, six pips at 0.0001 USD/EUR per pip. Each variant has a constant spread symmetric around the midpoint. Assume full Ask-to-Bid fills, USD 1.20 total commission, and no slippage or financing. A and B are invented variants, not observed sessions.

At a 0.8-pip spread in A, entry Ask is 1.10004 and exit Bid 1.10056: gross P/L is USD 10.40 and net USD 9.20. At a 2.4-pip spread in B, the prices are 1.10012 and 1.10048: gross USD 7.20 and net USD 6.00. The USD 3.20 difference isolates spread under identical midpoint movement. Spread is already in the fill-price result, so do not subtract it again. Real windows can differ in signals, price paths, fills and costs; a candle’s high-to-low range is not an automatically achievable profit.

The right hours depend on the pair you trade

Define Tokyo here as 09:00–18:00 Asia/Tokyo. This converts to 00:00–09:00 UTC, with different Warsaw clock hours in winter and summer. Yen and Australian-dollar releases may be relevant in that period; measure costs and strategy results rather than declaring it universally best or treating other pair movements as random.

Use the same instrument, account, quote size and sampling rules when comparing hours. Choose candidate windows according to relevant releases and your availability. The London and New York articles explain time-zone boundaries, while your records establish actual conditions.

Set exclusion conditions for your method

Identify conditions your method cannot tolerate before entry, including costs, missing quotes and execution uncertainty. There is no universal clock window in which costs necessarily consume every possible profit.

  • An Asian window can differ from a European one, but EUR/USD and GBP/USD are not inherently inactive then. Compare the actual spread records and range for your chosen account; a session name cannot establish that a scalping method has no opportunity.
  • Check your broker’s Friday closing time and holiday notices in their stated zone. For example, OANDA Corporation lists a Friday 16:59 New York close for most FX pairs (TRY pairs have a separate schedule). A universal 22:00 CET close is inaccurate; convert the published schedule for the date and consider weekend exposure.
  • Read your account financing cutoff. OANDA Corporation, for example, uses 17:00 New York time for overnight positions. This is 22:00 or 23:00 in Warsaw depending on the dated offsets. It is not a universal one-hour rollover window or proof of the widest spread.
  • Use a defined release rule. Record the publication’s dated source-zone time and check spread and execution conditions. For a hypothetical method that excludes entries above a 2-pip quoted spread, a 2.6-pip quote fails that rule whether or not news caused it. The threshold is a model assumption, not a universal optimum; waiting a fixed number of minutes does not guarantee normal execution.

Check holiday schedules and record conditions on the dates you actually trade. A month or season alone does not prove that liquidity is poor or the market is frozen.

Match the window to your day

Choose a window that fits your real availability, then test whether your strategy can operate under its measured costs and execution conditions. An after-work window is not automatically economical simply because New York remains within its local working hours.

Add dated releases to your calendar using the publisher’s zone and a conversion to your own zone. Select one or more windows deliberately, but review them with cost and strategy records. A fixed 14:30 European reminder will be wrong for some releases and clock-change periods.

Your next step

  1. Keep clocks for Europe/London and America/New_York alongside your own local zone. The two local boundaries and the date determine the overlap. One London clock alone does not account for the weeks when US and UK daylight-saving rules differ.
  2. Compare price range and transaction costs separately over several matched dates. A candle chart can illustrate movement, but Bid-only candles do not reveal the Bid–Ask spread or execution slippage. Keep the window definitions and observation frequency fixed.
  3. Record spread observations in each selected window and preserve timestamps, extreme readings and gaps. Three readings are a demonstration, not an estimate of lasting costs. Add commission and actual execution records before assessing transaction costs.
  4. Choose a window you can consistently supervise. If you use pending orders outside that window, check order acceptance, trigger rules, stop levels and execution risk. A pending order can be rejected or fill at a different price; it does not make a profitable plan automatic.

Use the trading sessions guide to define local windows and their dated intersections. Establish any hourly volume ranking from an appropriate timestamped data set rather than country turnover shares.