Why I Turn Off My Trading System During NFP
Calendar-based filter blocking entries around major economic releases, avoiding average 1.2R loss per major news event.
The Problem with News
Major economic releases (NFP, FOMC, ECB decisions, CPI) create instant volatility spikes that invalidate technical signals. The market's response to news is driven by the surprise relative to consensus, which no technical model can predict. Models trained on normal market dynamics will generate signals during news that look technically valid but are functionally random.
S29 blocks new entries in a 30-minute window around scheduled high-impact events. The calendar is loaded from economic calendar APIs at the start of each trading day. For unscheduled events (flash crashes, geopolitical shocks), S29 relies on S02's volatility filter and S10's autoencoder anomaly detection.
The 1.2R Average Loss
Analysis of trades that would have been entered during news windows showed an average loss of 1.2R. Not all were losers, but the distribution was sharply negative-skewed: the wins were small (market paused after news) and the losses were large (stopped out by the post-news spike).
Over the backtest period, S29 blocked approximately 120 signals around major news events. At an average 1.2R loss per event, that is approximately 144R in avoided losses, or about 19R per year. This makes S29 one of the highest-ROI modules on a per-intervention basis.
What Counts as "Major" News
S29 only blocks for high-impact events: central bank decisions, major employment reports, GDP, CPI, and PMI. Medium-impact events are not blocked because the average loss during medium-impact events is only 0.3R, which is within normal trading variance. The high-impact-only filter prevents overblocking. A system that avoids every economic release would miss 20% of all trading hours, far too many. Targeting only the releases with clear negative expected value keeps the filter surgical and the trade count healthy.