GreatFire and this site

Most of the data on this site's leaderboard wasn't measured by us. It is imported on a schedule from GreatFire's Circumvention Central. That's not hidden; every place imported data appears carries the line "Data source: GreatFire Circumvention Central". It is worth knowing who GreatFire is, how their data gets here, and where the two projects part ways.

Who GreatFire.org is

GreatFire is an anonymous group that has been active since around 2011, working on monitoring and pushing back against internet censorship in China. They first became known for the GreatFire Analyzer, a continuously updated database of which URLs and keywords are blocked in mainland China. When a news article says "site X is blocked in China", the evidence often traces back there.

Several projects followed. FreeWeibo collected and displayed posts deleted from Weibo. "Collateral Freedom" was a mirroring strategy: host mirrors of blocked sites on the domains of large cloud providers, so censors must either let them through or pay the price of blocking the whole cloud. AppleCensorship (later folded into an App Censorship project) tracked which apps Apple removed from its App Store in which countries. These projects vary in scale and current status (some active, some dormant), and their own site is the authority on the details.

Circumvention Central is one of their more recent projects.

What Circumvention Central does

The idea behind Circumvention Central (cc.greatfire.org, "翻墙中心" in Chinese) is direct: instead of relying on review sites' subjective scores, let real users inside censored networks run the tests and aggregate the results into a leaderboard. A user clicks a button in the browser, measures speed, latency and stability, selects a location and the tool in use, and the result goes anonymously into a public dataset.

Its leaderboard filters by region (mainland China, Russia, Iran, Turkmenistan, other) and by operating system, aggregates with medians, and requires a tool to have at least five tests spread across multiple days before ranking it. It also exposes a public API, which is what made this site possible.

GreatFire's approach has an obvious strength: the data comes from behind the wall. A speed measured by a reviewer overseas has little to do with what the same tool feels like in Beijing at 9 p.m. The weakness is equally obvious: the sample depends on volunteers, so popular tools get plenty of data while niche ones may see a handful of tests in months.

How this site imports

Our sync job runs an incremental import every hour at fifteen past, plus a full reconciliation daily at 03:30 UTC, pulling Circumvention Central's public data into our database. Four things come across: individual test results, the aggregates for each filter combination (used to compute the "GreatFire rank" column), the tool catalogue (including the service/self-hosted and free/paid labels), and each region's overall medians.

The fetching is deliberately polite. Requests identify themselves with a BreakHub-sync User-Agent that includes this site's address; all requests are sequential with at least 300 ms between them; failures back off and retry; only one sync can run at a time. The load we place on the upstream should be negligible.

Every imported result is tagged source = 'greatfire' and its id starts with gf_, distinct from the bh_ prefix on tests run here. In the interface, imported data carries a GreatFire source badge, and opening a single result shows a link back to the original upstream page (cc.greatfire.org/en/test?id=…). Every page that displays imported data ends with a "data source" note and the time of the last sync.

We don't alter imported data. Records the upstream deletes aren't proactively removed here, but full syncs reconcile them; records that are clearly bogus may be hidden (a hidden flag); hidden rows drop out of the statistics but are never deleted.

Comparing the two boards

Our ranking algorithm is the same as Circumvention Central's: average the per-metric ranks, and require at least five tests spanning 48 hours to enter the pool. On top of that we do two things.

First, we compute rankings over imported and locally run tests together, so our rank may differ from the upstream by a place or two. Second, the table keeps a "GreatFire rank" column, computed by running the same algorithm over the latest upstream snapshot. The two columns sit side by side so you can judge for yourself whether a difference matters. To see pure upstream data, switch the "source" filter to "GreatFire import".

There's a fuller explanation in How to read the rankings.

What we are not

A few things need to be said plainly.

This site is not affiliated with GreatFire. It is not a mirror or a regional branch, and it has not been endorsed by them. We use their public data and API, follow whatever usage norms we can observe, and credit them everywhere. If GreatFire asks us to stop importing, we'll stop.

This site does not accept paid placement, sponsorship slots or "recommended" labels from any VPN provider. Positions on the board are determined by data alone, and the algorithm is documented. Links in the tool catalogue point to providers' home pages with no tracking parameters attached.

Tests run here and imported data are labelled separately and never blended without attribution. Each local result carries full detail (per-URL speeds, per-domain latencies, per-image verification), while imported results only have the summary fields the upstream provides.

On privacy: we don't store visitors' raw IP addresses. Only when you've selected a circumvention tool do we resolve the exit IP to a country code and keep that. It's where the tunnel comes out, not where you are. The fingerprint lab stores only hashes and coarse attributes, never raw canvas or audio data.

What we add beyond the import

Circumvention Central is a dataset plus a leaderboard. On top of it we've added a reachability test (do ten commonly blocked sites open?), a three-stream peak speed, per-tool history pages, the side-by-side comparison of local and upstream data, and a separate browser-fingerprint lab. Chinese is the site's first language; English is a complete version, not a subset.

If you find Circumvention Central's data useful, go run a test at cc.greatfire.org; it will show up here within the hour. Running a test here works too; our data isn't sent back upstream, but it joins the upstream data in our rankings.

FAQ

Is this site official GreatFire? No. It's an independent project that uses their public data with attribution.

Do tests I run here get synced to GreatFire? No. Data stays on this site. To contribute upstream, test at cc.greatfire.org.

Why not just use GreatFire's site? You can, and we'd encourage it. What this site adds is Chinese-first copy, extra tests, history trends, the fingerprint lab, and both datasets in one view.

How often is the imported data refreshed? Incrementally every hour, fully once a day. The last sync time is at the bottom of each page.