When exactly do spreads start widening before a big release?

[Shadow Trader]

Not at the release. Earlier. Spread logs on a major pair show the widening beginning anywhere from one to five minutes ahead of a high-impact event, with most of the change in the last thirty seconds.

Liquidity providers withdraw size first, then widen. The platform's displayed spread may lag behind that, so the executable depth is already thinner than the quote suggests.

That timeline has a practical consequence for scalpers. A position opened at -2 minutes can be perfectly fine at the entry and then get closed at an exit that's three times as costly. Positions opened and closed entirely inside the pre-release window pay a bigger toll than they expected.

The exact pattern differs by event. Employment data and central bank decisions show the strongest effect. Mid-level data barely registers on some pairs.

Logging spread each second around ten releases on the pair of choice gives a plain picture. After a handful of events, the timing becomes predictable enough to set a "no new trades" cutoff based on evidence rather than guesswork.

When does the spread begin to widen on the pair traded before payrolls?

[FXS London Scalper]

Spread logs on a major pair show the widening beginning anywhere from one to five minutes ahead of a high-impact event, with most of the change in the last thirty seconds.

The last thirty seconds being where most widening happens matches what many London traders see before UK data. Cancelling pending orders at least one minute before avoids getting filled at a wide spread.

[Amanhasnoname]

On EURUSD with my broker the spread starts creeping up roughly two to three minutes before payrolls, but the real jump only comes in the final twenty to thirty seconds. I found this out the expensive way by holding a small scalp into the last minute because it was already in profit. The quote on screen still looked normal, yet the market order to close filled more than a pip worse than what I was looking at.

After that I wrote a tiny script that records the bid and ask every second, and I ran it through about a dozen releases. The pattern was steady enough to plan around. Payrolls and rate decisions were clearly the worst, CPI close behind, and most mid-tier data barely moved the spread at all on that pair.

My hard cutoff now is no new entries from five minutes before, and anything open gets closed by the three-minute mark whatever it is showing. I also pull pending orders, because a stop entry triggered by the spread alone is a horrible way to start the release.