Daylight saving time ends in the UK and the EU on Sunday 25 October, and in the US only on Sunday 1 November. For a trading bot, the trouble sits in the week between. From Monday 26 to Friday 30 October, most brokers' MT5 clocks run three hours ahead of London instead of two, and everything that comes out of New York lands an hour earlier on your clock. A bot that times its trades by a fixed server hour spends those five days just off the session it was built for.

Most bot owners never notice. Sometimes because the bot handles it properly. Sometimes because nobody looked. Ten minutes of checking is enough to know for sure.

Why does your bot run an hour off for a week?

Because your broker follows New York's clock and you follow London's.

Most MT5 brokers set their server to GMT+2 in winter and GMT+3 in summer. That is deliberate. Server midnight then falls exactly on 17:00 in New York, the moment the forex day closes, which gives you five clean daily candles a week with no stray Sunday stub.

Vantage, the broker we work with, does the same: GMT+2 as standard, GMT+3 while the US is on daylight saving time. In a notice about the spring change, Vantage said its server would move from GMT+2 to GMT+3 on Sunday 8 March 2026, with MT4, MT5 and the app down for maintenance between 09:00 and 13:00 server time. That was the American date. Europe only followed on 29 March.

So your server switches on the US date and your own clock on the European one. Normally the gap is a steady two hours: GMT+2 against GMT in winter, GMT+3 against BST in summer. During the in-between week the server is still on GMT+3 while the UK is already back on GMT. Three hours. And New York, normally five hours behind London, is only four behind from 26 to 30 October.

Which times move when daylight saving time ends?

Anything tied to New York comes an hour earlier. Anything tied to London or Frankfurt stays exactly where it was on your clock. On the MT5 clock it is the other way round.

Event Normal winter 26 to 30 October Summer time
London opens 08:00 08:00 08:00
London opens, on the MT5 clock 10:00 server 11:00 server 10:00 server
US data (8:30 New York) 13:30 12:30 13:30
Wall Street opens (9:30 New York) 14:30 13:30 14:30
Fed decision (14:00 New York) 19:00 18:00 19:00
Rollover and new daily candle 22:00 21:00 22:00

Every time is UK time except the second row, and the second row is where bots trip. A bot that catches the London open at 10:00 server time fires at 07:00 UK time that week. That is an hour before London opens, in the hour when only Frankfurt is properly awake.

News gets concrete too. Payrolls miss the window this year: September's numbers came out on 2 October and the next set lands in November. But the Fed decides on rates on Wednesday 28 October, and that hits UK screens at 18:00 rather than the usual 19:00. Weekly jobless claims on Thursday 29 October arrive at 12:30. How to trade around that kind of release is covered in our piece on NFP trading. That week, set the alarm an hour earlier. The Fed's meeting calendar is always in New York time, so do the conversion yourself.

Which bots are affected, and which are not?

It comes down to one question: does your bot look at the clock?

A bot that is allowed to open trades all day, with no time filter, barely notices the in-between week. On the Algo Trading Space forum someone asked how to adjust his EAs after his broker announced a move from GMT+3 to GMT+2. The answer was refreshingly dull. Trade all sessions and you change nothing. Trade one specific session and you compare the hours and adjust.

Where it goes wrong is EAs with fixed hours. In MQL5, functions such as TimeCurrent() and Hour() read server time. A setting like "trade from 10:00 to 13:00" therefore means 10:00 to 13:00 on your broker's clock, and that clock follows New York. A bot like that stays neatly aligned with America that week and slides an hour against London and Frankfurt. A breakout on the London open, like the London Open setup, is exactly the kind of strategy that suffers. The range it measures is a different range, and the breakout it waits for arrives an hour later than it expects.

A post on the MQL5 blogs from late August described the symptom plainly: a London box that was right on Friday sits one candle out on Monday. The EA is doing nothing wrong. It follows the times it was given to the letter.

The reverse happens too. An EA that works in GMT through TimeGMT() and applies the UK clock change itself stays locked to London and drifts against the US numbers that week instead. Working in GMT without that correction, it ends up an hour out on one side or the other after every clock change. London opens at 07:00 GMT in summer and at 08:00 GMT from 25 October. No single fixed offset keeps every session right at once. A well-built EA knows which session it is aiming for and handles daylight saving time itself.

What happens to your daily candle and your swap?

The MT5 daily candle opens at server midnight, which is 17:00 in New York. That holds all through the in-between week and you keep five daily candles. It just shows up at 21:00 on your clock instead of 22:00.

That matters in two places. Around rollover the market is at its thinnest of the day. Banks close their books and spreads widen, sometimes a lot on gold. Plenty of bots deliberately sit that hour out. If yours has that pause pinned to the local clock of your PC or VPS, say 22:00 to 23:30, it will keep trading that week right as spreads blow out.

The second place is swap. It is booked at rollover, and Wednesday night is the triple. On Wednesday 28 October the triple swap lands at 21:00 UK time, three hours after the Fed. If your bot closes positions to dodge swap based on a local hour, it will be late that evening.

Why doesn't your backtest show any of this?

Because the strategy tester reproduces the in-between week badly. According to MQL5's own documentation, TimeGMT() in the tester always equals the simulated server time. An EA that uses GMT live is quietly using server time in the backtest. The difference you want to test simply does not exist in the test.

It can surprise you from the other side as well. On the MQL5 forum, someone running a daily-candle EA in the tester saw it trade fine until a daylight saving date. Then on a Thursday it hit a "Market closed" error and never opened another trade. It had not happened with his previous broker. The moderator pointed him to the trading hours his new broker had set for that symbol.

If you want to know what your bot really did in one of these weeks, go to your live history. Filter on 27 to 31 October 2025, or 9 to 27 March this year, and lay the opening times of your trades next to a normal week. If they have all shifted by an hour, you know enough.

What to check before Monday 26 October

Ten minutes of work. Do it on Sunday night or Monday morning once the market is open again.

If you use someone else's bot, just ask: how does your EA handle the week between the European and the American clock change? A provider who cannot answer that straight away has probably never thought about it. That tells you more about the bot than one shifted hour ever will.

Should you switch your bot off that week?

No, almost never. An hour's shift on a bot that trades all day costs you less than a week of downtime. And a bot you switch off every time you feel unsure never runs long enough to prove its numbers.

Switching off only holds up for a strategy built entirely around one session hour, with fixed hours and no daylight saving logic, from a provider who cannot tell you how it works. Then five days of pause is cheaper than five days of trades at the wrong hour.

Spring is the bigger test, by the way. This year there were three weeks between the US change on 8 March and the European one on 29 March. In 2027 it is two, from 14 to 28 March. The exact dates per country are on timeanddate.com. Whatever you sort out now in October, you will not have to work out again in March.

Want a bot running on your own account?

You trade on your own MT5 account in your own name, with your own password, and you can stop whenever you like. We take 30 percent of your profit and nothing when you lose. Feel free to ask us how our bot handles the clock change.

Start now