Gateway leaderboard
Ranking algorithm rank v2: shrunk median minus its 90% confidence lower bound, multiplied by a critical-rate Wilson upper-bound penalty. Sites with fewer than 3 public samples are marked insufficient evidence in neutral gray — insufficient evidence ≠ problematic.
Three numbers, three different meanings:· Median score — typical performance over 90 days; · Rank key ⇅ — the only ranking input, blending performance, sample size and critical findings; · Critical-rate upper bound — the penalty term: any high_risk finding (model substitution, metering fraud, …) pushes the rank key down. Full methodology →
◎
No gateways listed yet. As soon as public detections exist, sites enter the board automatically under the same algorithm.
We never pad the board with unverified data.
Interactive comparison and certification badges are on the Chinese board. Site detail pages are published in Chinese; numbers are language-neutral.