A handful of scheduled releases move currencies, indices and gold more in a minute than they otherwise move in a day. The largest are the US employment report, US inflation figures, the Federal Reserve's rate decisions and the equivalents from the Bank of England and the European Central Bank. Below those sit GDP, retail sales, purchasing managers' surveys and central bank speeches, each capable of a sharp move on its day.
What happens at the release
Liquidity withdraws in the seconds before. Spreads widen, often to five or ten times normal. At the release, price jumps, sometimes in both directions within a second, and pending orders are filled at whatever price is available. Spreads then narrow over the following minutes as liquidity returns. A stop loss five pips from the price is almost certain to be hit during this, whether or not the eventual move was against the position, because the spread alone can reach it.
The calendar
MetaTrader 5 includes an economic calendar in the Toolbox, showing each release with its expected impact, and most brokers publish one. Events marked high impact are the ones that matter. Knowing when they are is enough to avoid opening a trade into one, and it is the basis of every news filter.
At 1.30pm UK time on the first Friday of the month the US employment figure is released. EUR/USD is at 1.0850 with a one-pip spread. Ten seconds before the release the spread is nine pips. At the release the price prints 1.0872, then 1.0838, then settles around 1.0860, all within five seconds. A trader long from 1.0850 with a stop at 1.0840 is stopped out at 1.0836, four pips of slippage beyond the stop, on a move that ended ten pips in their favour. A trader who was flat and waited until 1.45pm entered at a one-pip spread into a market that had decided its direction.
An EA has two honest choices. Trade through news and accept that a few trades a year will be stopped out by the spike, which is acceptable for strategies with wide stops on higher timeframes, where the spike is a small fraction of the stop. Or block new entries for a window around high-impact releases, which is the sensible design for anything with tight stops or short holding periods. Many EAs offer the filter as a setting; on funded accounts, where some firms forbid trading around news, it is usually mandatory.
A backtest cannot model any of this, because historical data does not show the spread widening or the fills. A live record does. Look at how an EA's live trades behaved on the release days, since that is where the difference between a backtest and reality is largest.