Session Trading vs All-Day Trading: One Rule, 24 Hours, 3.55 Pips of Difference

Most traders sit at the chart for as long as they are awake and take whatever appears. The alternative is to pick a few hours and ignore the rest. I have a way to settle that argument with numbers rather than opinion: run one identical mechanical rule at every hour of the day across 16,926 EURUSD 5-minute bars, and see how much of the profit comes from how little of the clock.

What exactly is being compared?

One rule, twenty-five starting points. The rule is the same release-candle breakout from my news release test, but stripped of any connection to news: take the opening 5-minute candle of an hour, mark its high and low, enter on the first break within 15 minutes, fixed stop and target, hard exit after one hour.

• All-day version: the rule fires at the top of every hour, 24 times a day.
• Session version: the rule fires only in a chosen window.
• Data: 16,926 EURUSD 5-minute bars, June 3 to August 26, 2026 — the same set behind my Ichimoku cloud breakout test and my ADX test.
• No overlap: 1,193 trades in total. Most hour slots carry around 50, but the rollover hours are thinner — 21 trades at 22:00, and 41 each at 21:00 and 23:00 — so those rows are weaker evidence than the rest of the table.

Because only the starting hour changes, any difference between columns is a property of the clock and nothing else.

Which hours actually pay?

The table shows 15 of the 24 hourly slots — the best, the worst and the boundaries; the omitted rows sit between +0.2 and +1.6 and change nothing about the ranking. Note also that the samples are not equal: most hours carry around 50 trades, but 22:00 has only 21 and 21:00 and 23:00 have 41 each, because the rollover hours produce fewer clean breakouts. The bottom rows are the thinnest evidence on the page.

The spread between the best and worst hour is 3.55 pips per trade on an identical rule. That is larger than the difference between most of the indicators I have tested against each other, including the whole four-way Ichimoku against trend following comparison. The clock moved the result more than the indicator did.

And 01:00 UTC beating everything is not what the textbooks predict. That hour sits in the middle of Asian trading, the session usually described as too quiet to trade. It is quiet — but a quiet hour where a breakout resolves cleanly, with few false breaks to whipsaw through, turned out to be the most reliable slot on the board at a 67.9% win rate.

Is the news window worth the reputation?

At the headline number, yes: +2.13 against an all-day +1.08. But test how it survives a change of settings and it looks different.

Read the 10/20 column. The news window loses three quarters of its edge the moment you ask for a wider move, because a release is a burst that ends. The 01:00 window holds +2.62 with the wide setting, almost identical to its tight one, because whatever it is capturing keeps going. Two setups with similar headline numbers and completely different characters.

That distinction matters more than the ranking. A setup that only works at one exact stop and target is a setup you have fitted; a setup that pays across three different settings is one that describes the market. On this data 01:00 and 08:00 are the robust hours and 12:30 is the delicate one.

What does trading all day actually cost?

Take the eight best hours and you average roughly +2.1 pips per trade. Take all twenty-four and you get +1.08. The other sixteen hours are not free — they are actively cutting your result in half, and two of them are paying you to stay away.

There is a second cost the profit column hides: the spread. In the quietest hours the average 5-minute candle is under 1 pip tall, so a standard 1 pip spread is larger than the whole bar. I break that down hour by hour in my guide to spread and swap costs on MT4, and it changes which of these hours are tradable in practice rather than on paper.

What I would actually do with this

• Pick hours, do not sit all day. The same rule doubled its per-trade result by dropping two thirds of the clock.
• Do not assume the busy hours are the good ones. 01:00 UTC beat every US hour on this data.
• Cut 18:00 and 20:00 UTC entirely. Both were negative on a rule that made money nearly everywhere else.
• Prefer the hours that survive a settings change. 01:00 and 08:00 held up across 5/5, 5/10 and 10/20. The news window did not.
• Check the spread before you trust an hour. A positive result on mid-price can be negative after costs when the candle is under a pip.

Frequently asked questions

Is session trading better than trading all day?

On this data, clearly. An identical breakout rule made +1.08 pips per trade averaged across all 24 hours and +2.13 to +2.65 in the best individual hours. The gain comes from removing losing hours rather than from finding better entries.

What is the best hour to trade EURUSD?

01:00 UTC had the strongest result here — 53 trades, 67.9% win rate, +2.65 pips per trade — followed by 08:00 and 06:00 UTC. This measures one specific breakout rule on one three-month sample, so treat it as a place to start testing rather than a law. My breakdown of the best hours to trade EURUSD approaches the same question from the volatility side.

Why were 18:00 and 20:00 UTC losing hours?

Both sit after the US data flow has finished, when liquidity thins but the day’s range is already set. Breakouts in that stretch tend to be the tail end of moves that are done, and the win rates show it: 38.5% and 32.7%. The 18:00 slot also had the worst continuation figure in my news release test, at 13.6%.

Does this apply to other pairs?

The method does, the specific hours may not. EURUSD’s clock is shaped by London and New York; a yen or Australian dollar pair has a different centre of gravity. Run the same hour-by-hour split on your own pair before adopting these numbers.

How many trades is this based on?

1,193 trades in total. Most hour slots carry around 50, but 22:00 has 21 and 21:00 and 23:00 have 41 each. All from 16,926 EURUSD 5-minute bars between June 3 and August 26, 2026 — the same dataset as my stop loss distance test and my moving average crossover test.