Trading Sessions & Timezones in Backtesting

The Session Map (in Server Time)

On the standard GMT+2/+3 feed, sessions land in fixed server-time slots — that’s the whole point of anchoring there:

Session	Server time (approx)	Character
Sydney/Asia	00:00–10:00 (active 02:00–10:00)	Thin early, range-prone; JPY/AUD most active; widest spreads near rollover
London	10:00–18:00	Deepest FX liquidity ramp; trends and breakouts; tightest spreads begin
London–NY overlap	15:00–18:00	Highest volume of the day; major moves and reversals cluster here
New York	15:00–23:00	US data prints (15:30 server typical); liquidity fades after 20:00
Rollover	~23:55–00:15	Spread spikes, swap booking, avoid resting tight stops

Index CFDs differ: they follow their underlying cash market’s hours with thinner synthetic pricing outside them — a Nasdaq 100 strategy backtested in cash hours and traded overnight is two different strategies (see gaps & rollovers).

Why Session Context Belongs in the Backtest

Three of this site’s recurring findings are really session findings: spreads are session-dependent (the cost audit’s strategy died when cloned to Asia purely on costs), volatility is session-dependent (a 25-pip stop is wide in Asia, tight in the overlap), and behavior is session-dependent (breakouts that run in London fade in Asia). Minimum implementation: log the session of every backtest trade — one journal column — and read your expectancy segmented by session before trusting it. Strategies often turn out to be one-session edges wearing an all-day costume; that’s not a flaw, it’s a free regime filter.

Writing DST-Proof Session Rules

The naive rule “London opens at 8:00” is anchored to a wall clock somewhere. Three habits make session logic survive clock changes:

1. Anchor to server time (after detecting the data’s offset) — on a US-DST server, US-anchored events stay fixed (NY open stays 15:30 server year-round) while London-anchored ones drift only during the 2–3 week US/UK DST divergence windows.
2. Prefer behavioral anchors over clock anchors where possible: “the first M15 range after the London liquidity ramp” self-adjusts; “08:00 sharp” doesn’t.
3. Flag the divergence weeks. Mid-March and late October backtest trades deserve an asterisk — if a session strategy’s losers cluster there, you’ve found the clock, not the market.

Related: GMT & DST facts · spread by session · testing across regimes

Frequently Asked Questions

What are forex sessions in plain GMT?

Approximately: Sydney 21:00-06:00 GMT, Tokyo 00:00-09:00 GMT, London 08:00-16:30 GMT (07:00-15:30 in UK summer), New York 13:30-22:00 GMT (12:30-21:00 in US summer). The drift in those parentheses is exactly why anchoring backtest rules to a consistent server clock beats quoting GMT.

Which session should a beginner backtest first?

London and the London-NY overlap: deepest liquidity, tightest spreads, most directional behavior — and therefore the most forgiving cost structure for a developing strategy. Asian-session strategies are viable but live or die on spread discipline, as the cost numbers show.

Do sessions matter for swing trading backtests?

Less for entries, more for management: daily candle boundaries (rollover), swap booking times and gap windows are all session-clock events that touch multi-day positions. A swing backtest mainly needs the right daily-candle anchoring rather than intraday session logic.

How do I handle sessions for index CFDs?

Anchor to the underlying cash market's hours (e.g. US indices: 16:30-23:00 on a GMT+2/+3 server, shifting with US DST), and treat out-of-hours CFD pricing as its own thin regime — wider spreads, gappier moves. Backtest in-hours and out-of-hours behavior separately; they rarely belong in one statistics pool.