Timing·6 min read

Trading sessions

In one sentence
The hours during which a major financial centre is open, used to divide a 24-hour market into periods that behave differently.

Also called Killzone, London open, New York open, Asian range, Session overlap — all covered here rather than on pages of their own.

Published 15 September 2026 · by the SageTradingJournal team

Spot forex has no exchange and no closing bell, but the people trading it keep office hours. As Tokyo, then London, then New York come online and go home, the number of active participants changes, and with it the amount of movement in a typical hour. Dividing the day into sessions is a way of admitting that 03:00 and 13:00 are not the same market.

SessionRoughly (UTC)Usually described as
Sydney21:00 – 06:00The quietest; the start of the trading day
Tokyo / Asia00:00 – 09:00Ranges more often than it trends; sets levels for later
London07:00 – 16:00The largest by volume; most daily ranges begin here
New York12:00 – 21:00Where scheduled US data lands
London / New York overlap12:00 – 16:00Both centres open at once

Those hours are approximate and they move. London and New York shift with daylight saving on different dates, which means the overlap is four hours for most of the year and three for a couple of weeks twice a year. Any session rule stated in your own local time will quietly drift by an hour, twice a year, in a way that is nearly invisible in a backtest and completely invisible in memory.

Timezones are the hard part

Before any claim about sessions can be tested, one boring question has to be answered: what timezone are the timestamps in? Most MetaTrader brokers stamp candles in a server time that is neither UTC nor your own, often shifting with a daylight-saving rule belonging to yet another country.

What "killzone" adds, and what it does not

A killzone is a narrower window inside a session — commonly the first hour or two after a major centre opens. The underlying observation is unremarkable and well supported: activity concentrates around opens, and around scheduled data. More participants and more scheduled events in an hour means more movement in that hour, on average, than at 03:00.

More movement is not the same as more opportunity. A wider range is available to be captured and also available to run through a stop, spreads are frequently wider around opens and releases, and slippage is worst exactly when movement is fastest. Whether a busy hour is a good hour for you depends on your entries, your stop distance and your costs — which is a question about your record, not about the market.

What your own record can tell you

This is one of the few concepts in trading where a personal answer is genuinely within reach, because every trade already carries a timestamp. No tagging, no extra discipline, nothing to remember: the data is there from the first trade you ever logged.

  • Expectancy by session, in R, with the number of trades in each bucket — the count matters more than the average when a bucket holds six trades.
  • Expectancy by hour, which often shows something a session average hides: one bad hour dragging an otherwise fine session down.
  • Where your losses cluster. If they concentrate in one window, that is worth knowing before any conclusion about which hour is best.
  • Cost by hour. Spread and slippage vary through the day, and a busy hour can be worse after costs than a quiet one is before them.

Sage's analytics break results down by session and by hour from the timestamps already on your trades, and the backtesting workspace marks session boundaries on the replay chart so a rule about an hour can be tested rather than remembered. Both report counts alongside the numbers, because a session that looks wonderful across nine trades has not told you anything yet.

A caution about slicing

There are twenty-four hours, several sessions, five weekdays and any number of ways to combine them. Search hard enough through a modest record and something will look excellent — Tuesdays in the London open, say — for no reason other than that you looked at forty combinations and kept the best one.

The protection is ordinary and slightly tedious: decide which split you are testing before you look, keep the sample sizes visible, and treat a finding from a slice you discovered by searching as a hypothesis for the next hundred trades rather than a conclusion about the last hundred. The guide to statistics that matter goes through how thin a slice gets before it stops meaning anything.

Where traders disagree

This vocabulary is not standardised. These are the live disagreements — worth knowing about before comparing your results with anyone else’s.

  • Where each session starts and ends — published hours differ by several hours between sources, and each choice changes which trades land in which bucket.
  • Whether "killzone" windows describe anything beyond the ordinary concentration of activity around opens and scheduled releases.
  • Whether session effects persist, or drift as market participation and automation change over the years.
  • Which timezone a session rule should be stated in — broker server time, UTC, or the local time of the centre concerned.

Test it on your own trades

Whether this holds is a question about your record, not about the term. These are the steps that make the answer trustworthy.

  1. 1Establish your data's timezone offset against a known event before bucketing anything; an hour of error invalidates the whole breakdown quietly.
  2. 2Read expectancy by session and by hour with the trade count shown beside every figure.
  3. 3Check the cost side too — spread and slippage vary through the day and can reverse the ranking.
  4. 4Decide the split before looking at the result, or treat what you found as a hypothesis for the next sample rather than a conclusion about this one.
  5. 5Re-check the finding on a later period; session effects are among the easiest to discover by accident and the hardest to reproduce.

You will also need

Questions, answered.

What are the best hours to trade?
There is no answer to this in general, and anyone offering one is describing their own record or nobody's. Your trades already carry timestamps, so the question can be answered specifically for you — expectancy by hour, with the count beside it, which is a far more useful number than a published table.
Why do the session times differ everywhere I look?
Because they are conventions rather than facts, they are quoted in different timezones, and daylight saving moves London and New York on different dates. Pick a definition, state the timezone explicitly, and keep it fixed across the whole sample.
What timezone is my broker using?
Often a server time that matches neither UTC nor your own, sometimes with its own daylight-saving rule. Confirm it against a scheduled release or the daily candle boundary rather than assuming — a silent one-hour offset is the commonest reason a session finding cannot be reproduced.
Does the Asian range matter?
Some traders use the overnight range as a reference level for the sessions that follow. That is a rule, and it is testable like any other: define the range, define what counts as a reaction, and count both the occasions it held and the ones it did not.

Stop believing. Start counting.

Tag the trades that used this, and Sage reads back what they actually did — in R, against your own baseline, with the sample size attached. It will not tell you the pattern is good; your record will.

Free · no card · your data stays yours