The independent benchmark
for residential proxies.

Probes run every 15 minutes, around the clock, from EU and US vantage points. Results are aggregated once a day into a daily rollup, and the composite score is recomputed at that point — so leaderboard numbers change once per day, not per probe. Trend charts plot this real daily history over 7-day, 30-day, and 90-day windows; nothing is projected or interpolated.

Yes. ProxyStats is not owned or operated by any proxy provider. We test standard retail proxy packages — some purchased, some granted by providers as evaluation traffic — routed exactly like a regular customer's workload. Access to testing never affects scores or placement: the same automated probes measure every provider, and no one can pay for a better position. Every score is reproducible from the published methodology.

We sample live Google and Amazon endpoints on a slow schedule with strict per-IP cooldowns. A CAPTCHA, an interstitial challenge page, a 403, or a timeout all count as failures — an HTTP 200 that serves a challenge instead of content is not a success. One honest note: this pass rate is saturated among credible residential providers, so we show it but rank use-cases by what actually separates pools — latency, session stability, and clean-IP share.

A lower success rate usually means a network has a high concentration of datacenter-flagged IPs or a poorly optimized routing pool. Proxies that are heavily abused by botnets shrink the clean IP pool, so target servers block the remaining connections more often.