Event-Drift Live Report 2026-W31 — every trade we actually placed on macro surprises

What this is

This is not a market-forecast article. It is the record of trades we actually placed.

There is one method. We measure how far an economic release landed from its forecast (the surprise, in standard deviations), enter in that direction on the first hourly bar to close after the release, and exit after 24–48 hours. There is no take-profit. The stop is 3× the average range (ATR).

No chart patterns are used. Central-bank policy events are deliberately excluded — in back-testing they ran a 31% win rate in the later period.

This week

No qualifying release occurred this week, so no trades were taken. Only a narrow set of indicators can trigger this system, so quiet weeks are normal.

Cumulative record

No closed trades yet.

R is profit divided by the amount risked on that trade. We report R rather than currency because it compares the method itself, independent of account size or denomination. As a practical benchmark, averaging above +0.2R over a large sample means the system is working.

Measured execution (the part nobody else publishes)

Not measured yet — no entries so far.

Broker measurements

These figures come from placing real minimum-size orders on both brokers at the same time and measuring the fills (around 07/28 21:28 JST). The margin row compares what MT5 reports as required margin against what was actually held. On Exness, five times the API figure was locked up (reproduced across two fills). Sizing from the API value alone would leave you larger than intended.

So which one should you open?

Different axes have different winners. Reading this as "which is better overall" misses the point.

• On trading cost, Exness. The measured USD/JPY spread differs by roughly 3.7×, and that applies on every round trip. The more you trade, the more it compounds.
• On the account reporting truthfully, XM. Its required-margin API matched reality. Exness was off by 5×, which is a practical trap if you size positions programmatically. If you size by hand, it does not affect you.
• On trying it without depositing, XM. The no-deposit bonus is on that side.

We hold accounts at both and placed real orders at both to produce these numbers. The spread gap is large, but it is not on its own a reason to rule either out — the answer depends on which of the three axes matters most to you.

Releases we are watching next

• 07/30 21:30 JST — USD Advance GDP q/q (4 pairs)
• 07/30 21:30 JST — USD Core PCE Price Index m/m (4 pairs)

What we will say plainly

• A ~45% win rate is the design, not a failure. With no take-profit and a time-based exit, we do not win often; the edge is in how large the winners are. Reading the win rate alone will mislead you.
• Losing streaks are certain. For any positive-expectancy system at ~45%, runs of 5–7 losses are statistically ordinary.
• Back-testing was negative through the low-volatility 2018–2022 regime. We tested five regime-detection methods to avoid that period; none had predictive power. We disclose it as accepted risk.
• Results are execution records from a demo account. Past results do not guarantee future performance. This is not investment advice.