Your broker's clock is not your clock

Every time filter in every EA is written in server time. Not your time, not London time, not the time in the tutorial you followed. Get the offset wrong and a bot built to avoid the first hour of the New York session trades right through it, and the equity curve slowly stops matching the backtest for a reason nobody thinks to look for.

Where the offset comes from

Brokers choose their own server time zone. Many sit at GMT+2 in winter and GMT+3 in summer, which places the daily candle close at 5 pm New York time and gives a clean five-day trading week. Others run GMT+0, or GMT+3 all year, and a few sit somewhere unusual. The result is that the same strategy, on two brokers, is running two different schedules.

Check it, do not assume it

Market Watch shows the server time above the symbol list. Compare it to your own clock and write the difference down. In India that is usually IST minus two and a half or three and a half hours, but only usually, and only until the next daylight-saving shift.

Daylight saving moves the market, twice

Europe and the United States change clocks on different weekends, so for a couple of weeks each year the gap between the London and New York sessions is an hour different from normal. If your broker follows one of them and your filter was tuned to the other, your session window drifts. Nothing errors. The bot simply trades a slightly different market than the one it was tested on.

Twice a year, every time-filtered strategy is running an hour off. The fix is a calendar reminder, not a code change.

Five candles or six?

A quick way to identify a server's time zone: count the daily candles in a normal week. Five means the daily close lines up with the New York close, which is what most gold and forex strategies are built around. Six means Sunday gets its own short candle, and any indicator using daily values — pivots, previous-day high and low, ATR — is calculating from different data than the same indicator on a five-candle broker.

Neither is wrong. But a strategy tuned on one and run on the other is not the same strategy.

Backtests inherit the same problem

The Strategy Tester uses the history from the broker you are logged into, in that broker's time. Take the results to a different broker with a different offset and the session filters no longer point at the same hours. When you switch brokers, re-run the test on the new server before you trust the old numbers.

What we do

The server offset goes into the same note as the account number and the magic numbers. When something behaves oddly in late March, the first question is whether the clocks moved, and it takes ten seconds to answer instead of a weekend of theories.