News volatility around the NFP: why it cannot be reproduced
Methodology
- The hypothesis requires the exact release time and the intraday price dynamics in the first minutes after publication.
- Daily bars give one point for the entire day and contain neither the event marker nor the behaviour of the spread at the moment of the news.
Why the hypothesis cannot be tested on our data
The hypothesis is tied to the event calendar and to the first seconds after the NFP release, with a sharp spread widening. Daily bars contain neither event marks nor intraday execution dynamics — a check on closing prices would be flattering.
We deliberately give no figures here: a result computed on unsuitable data would be a fabrication. Blank space is more honest than a fake study.
Download the exact sample and run the logic yourself — the numbers above should match.
A Binance spot EUR/USDT proxy series, not a forex-broker feed. Binance Spot REST API (api.binance.com/api/v3/klines).
Conclusion
The hypothesis cannot be tested on daily data: the outcome of news trading is determined by the first seconds after the release, when the spread widens sharply. That information is not in a daily bar, and a backtest on closing prices would hide the execution costs.
Practical takeaway for the trader
News trading is determined not by the forecast but by execution at the moment the spread widens. Judging it without intraday data and a slippage model is pointless.
FAQ
Why can it not be tested on daily bars?
The outcome is determined by the first minutes after the data release, when the spread widens several times over. A daily bar gives one point for the day and contains none of that dynamics.
What would a closing-price backtest hide?
The execution costs at the moment of the news — the spread widening and slippage. It is those, not the choice of direction, that most often determine the result.
What would be needed?
An event calendar, tick data around the events and a dynamic-spread model — without them a check of the news hypothesis is inherently incomplete.