How it works
How the trading day is shaped
The FX week opens in the Asia-Pacific region on Sunday evening, New York time, and closes on Friday at 17:00 New York time. In between, liquidity rises and falls as regional centres open and close. Rough hours in UTC (they shift by an hour with daylight saving):
| Session | Approx. hours (UTC) | What tends to happen |
|---|---|---|
| Sydney | 21:00 to 06:00 | Thin, especially in the first hour after the weekly open |
| Tokyo | 00:00 to 09:00 | JPY and AUD pairs more active, EUR crosses quieter |
| London | 07:00 to 16:00 | The deepest session for most pairs |
| New York | 12:00 to 21:00 | US data at 12:30 or 13:30 UTC; overlap with London is the busiest stretch |
The London and New York overlap is where spreads on the majors are usually tightest. Spreads are at their worst around the 17:00 New York rollover, when many banks close their books, liquidity providers step back for a few minutes, and spreads widen sharply. News releases cause the other big spikes.
Worked example (illustrative numbers)
Say a strategy trades EURUSD 200 times a month, each round trip paying the spread once.
- If you assume a flat 0.2 pip spread, cost is 200 × 0.2 = 40 pips a month.
- Now suppose 30 of those trades fire in the hour around rollover at an average of 1.5 pips, and the other 170 trade during the overlap at 0.2 pips. Real cost is 30 × 1.5 + 170 × 0.2 = 79 pips.
The flat assumption hid almost half the cost. For a strategy that targets 5 to 10 pips per trade, that is the difference between a profit and a loss.
Keep reading
Measure spread by hour from tick data
Tick data with bid and ask lets you build a spread profile for each pair. In our FX-TICK schema every row has symbol, timestamp, bid, ask and mid.
"""Example: median EURUSD spread in pips, by UTC hour."""
import pandas as pd
ticks = pd.read_csv("eurusd_ticks.csv", parse_dates=["timestamp"])
pip = 0.0001 # use 0.01 for JPY pairs
ticks["spread_pips"] = (ticks["ask"] - ticks["bid"]) / pip
ticks["hour_utc"] = ticks["timestamp"].dt.hour
profile = ticks.groupby("hour_utc")["spread_pips"].agg(
median="median", p95=lambda s: s.quantile(0.95), ticks="count")
print(profile.round(2))
Read three things from the output. The median column shows the normal cost per hour. How bad it gets shows up in p95, and around rollover it is often several times the median. Activity is in ticks: hours with few ticks are hours where quotes go stale and your fills will differ from the backtest.
If you only have 1-minute bars, look for an average_spread field, which our FX-BARS schema includes. It is coarser than ticks but much better than a fixed number.
Watch out
Common mistakes
One spread for the whole day. The cheapest hour and the most expensive hour can differ by ten times or more. Use the hourly profile, or at least separate rollover and news windows.
Backtesting on mid prices with no spread at all. Many free FX files are mid only. A strategy that trades often can look excellent on mid and lose on bid and ask.
Ignoring the Sunday open. Weekend news gets priced in when the market reopens, often with a gap and very wide spreads for the first minutes. Stops placed over the weekend can fill far from their level.
Forgetting daylight saving. London and New York switch clocks on different dates in spring and autumn. For a week or two the overlap is an hour shorter or longer. Store timestamps in UTC and convert per date.
Details
Get the data
FX ticks with bid and ask, so you can measure spreads yourself. What you'll get:
bid,ask,midand sizes where the source provides them, timestamps in UTC to the millisecond.
- Forex tick data, available in early access. Suggested price: $3 per pair per month of ticks; one year of ticks is planned for Pro ($79/mo).
- Forex intraday data adds
average_spreadto every bar.- A free FX tick sample is available on request. Our free ECB reference rates sample is daily only (source: European Central Bank).
FAQ
Frequently asked questions
Which forex session has the lowest spreads?
For the major pairs, spreads are usually tightest during the London and New York overlap, roughly 12:00 to 16:00 UTC. JPY and AUD pairs can also be tight during the Tokyo session. Always check with tick data for your own broker or source.
Why do forex spreads widen at 17:00 New York time?
That is the daily rollover, when banks close their books and many liquidity providers pull or widen quotes for a few minutes. Spreads normally return to their usual level shortly after.
Do forex sessions matter for a backtest?
Yes. Spread, volatility and fill quality all change by session. A backtest that ignores them tends to overstate results for strategies that trade often or near the rollover.
How do I calculate the spread in pips?
Subtract the bid from the ask and divide by the pip size, which is 0.0001 for most pairs and 0.01 for JPY pairs. For example, a quote of 1.09680 / 1.09682 on EURUSD is a 0.2 pip spread.
Keep reading