News Filters in a Trading Bot: Why They Must Block Entries but Never Exits

Among all the dangers facing an automated strategy, high-impact economic releases are the only fully predictable one: the time is known days in advance. That's exactly why they belong in a bot's logic - and exactly why our Opening Range Breakout bot has had such a filter from the start.

Why a backtest barely shows the problem

In historical data, news timestamps often look unremarkable. What's typically missing is precisely what causes the damage live: abruptly widened spreads, thin order books, fills far away from the expected price. A backtest still fills the order cleanly at the desired price, painting a picture around such events that is considerably friendlier than reality.

The most common implementation mistake: blocking everything

The first draft is almost always: during the news window, the bot does nothing at all. That's more dangerous than having no filter. Because if a position is open when the window starts, a blanket block means stop trailing, break-even logic, and exits all stand still exactly when the move is most violent. The correct split: the filter blocks new entries only. Everything that protects or closes an existing position keeps running unchanged.

Relevance instead of a blanket block

The second typical oversimplification: every calendar event gets treated the same. That produces unnecessarily long downtime for events that don't affect the traded instrument at all. What makes sense is a mapping - which instrument depends on which economy, and does the event belong to it? Only when both line up is a block justified.

What should stay configurable

• Windows before and after the event, set separately - the trailing window often matters more than the leading one
• An impact level, so only genuinely market-moving releases block and not every minor print
• Behavior with an open position: leave it running, close beforehand, or only tighten the stop - a deliberate decision rather than a silent default
• Logging of blocked entries, so it can be judged afterward whether the filter was drawn too tightly or too loosely

The general limits of automated strategies are covered in more detail in our Risk & Liability Notice.

A news filter isn't protection against losses - it's a deliberate decision about which predictable events a strategy should take part in at all. The difference between a useful and a harmful filter almost always comes down to what else it switches off.