执行质量报告：测量毫秒

This explainer uses an illustrative venue dataset — 4.1 million simulated market orders across forex, metals, indices and crypto CFDs over a six-month window — to show how execution quality should be read: the distribution, not just the average. Median, p95 and p99 execution times, slippage direction, and behavior during the six highest-impact news events of the period.

Averages hide tails. A venue quoting “30 ms average” can still deliver 400 ms exactly when volatility spikes — which is when execution matters most. Percentiles tell the truth instead.

The p99 figure includes the release minutes of every NFP, CPI and FOMC event in the window. Even the worst 1% of orders completed in under an eighth of a second.

Slippage direction is the honesty test for any execution venue. Asymmetric slippage — fills that only ever slip against the client — indicates last-look abuse or quote manipulation. A venue handling slippage symmetrically produces a split like the example below:

If a venue won't publish its negative-slippage share, assume the number answers the question.

— 实用的经验法则

The example isolates the 60-second windows around six tier-1 releases. Spreads widen — that is real liquidity thinning, not venue behavior — but execution time and reject rates stayed inside normal bands:

• Median latency during release minutes in the example dataset: 34 ms (vs 28 ms baseline).
• Reject rate: 0.4% (vs 0.2% baseline) — rejects re-quote rather than silently fill at a worse price.
• XAUUSD spread at NFP peak: 3.2× typical, recovering to baseline within 41 seconds on average.

• Population: every market order on live accounts, Jan 1 – Jun 30 2026. No sampling, no exclusions except cancelled-before-fill.
• Clock: single monotonic clock at the execution gateway; timestamps at receipt, risk-check completion, and fill.
• Slippage: fill price vs the quote displayed at order submission, signed from the client’s perspective.
• News windows: ±30 s around the scheduled timestamp of NFP (×2), US CPI (×2), FOMC (×2).