How the ping test measures

What the test times

When you tap Run, your browser opens a connection to the test point of each region with one warm-up request. The warm-up pays for the address lookup and the secure handshake, and it is not counted. Then the browser sends 10 small requests over that open connection. Each one starts at least 0.3 seconds after the previous one started, and the test points start 0.05 seconds apart, so all 20 regions run side by side and the test takes about 10 seconds.

A request is timed from the moment the browser sends it to the moment the answer starts to arrive. A request with no answer within 2 seconds, or with an answer other than a plain OK, counts as lost; the warm-up gets 4 seconds. The whole test stops 12 seconds after the tap. A request that has not been sent by then is not counted as lost, and the result says how many did not run. A lost request usually closes the connection, so the browser opens it again with one more untimed request before the next timed one. A test point that stays silent through its warm-up and its first request stops there and reads not reachable.

The test runs only while its tab is visible. Browsers slow down timers in a hidden tab, so switching away stops the run, and the page asks you to run it again.

The figures for each test point

• Median. The answered times, sorted, and the one in the middle. With 10 answers it sits halfway between the 5th and the 6th. This is the figure a result prints.
• Spread. The width of the middle half of the sorted times: the time a quarter of the way up the list subtracted from the time three quarters of the way up. One slow request barely moves it; a line that keeps jumping widens it.
• Lost. 10 minus the requests that got an answer.
• Not reachable. A test point with fewer than 5 answers. It gets no figure, it never counts as the nearest region, and the result never fills in a guess.

The test works with tenths of a millisecond and prints whole milliseconds.

The bands

A test point is unstable when its spread is over 20 ms, or when 2 or more of its 10 requests got no answer. A spread of exactly 20 ms, or 1 lost request, still counts as steady.

From test points to a region

Some regions have more than one test point; Europe West has three. A region takes its best test point in this order: a close and steady one first (Excellent or Good, not unstable), then a close one that jumps, then the rest, and inside each group the lower median. A failed test point is skipped.

Publishers often name an area without a city, such as Europe or Asia. The test then measures the regions inside that area and takes the best one by the same order, and the game page prints the publisher's own label with city not published: an area never borrows a region's city. The data page lists every region and its test points.

Which region the verdict reads

1. The region you picked under Region you play on, or the server you named on a game that ties you to one.
2. Without a pick: the region a publisher says players from your country land on, when that region has a test point. The data holds two such statements: NARAKA: BLADEPOINT for Vietnam and Roblox for Brazil.
3. Otherwise the best region by the order above, the way a game that picks the server for you would. A steady region at 44 ms wins over an unstable one at 38 ms.

For a game whose publisher lists its regions in full, the test compares that game's regions only. For a game with a partial list or none, and for the test without a game, every region counts. The nearest region is the answered one with the lowest median among the regions that count.

A result speaks about a game's own servers when the publisher's list is complete, when you picked the region, or when a published country default applied, and always on the test without a game. A game with a partial list or none, and no pick, gets a general result, and the result says so.

The four verdicts

A region that tests Good stays Close and stable even when a nearer region is faster: a good ping on a far region is no reason to buy anything. A worse measurement never makes a verdict better; the test checks that rule on every worked example in its own test suite.

A link to a route optimizer appears only on Far from the servers and Cross region, only on a result about the game's own servers, and only with the VPN box unticked. A close result never shows one. How we make money explains the link, and the GearUP Booster page what that product says it does.

Results without a verdict

• No test point answered. A content blocker, a firewall or a Wi-Fi sign-in page can stop every request. The page shows no figure and asks you to check the connection and run again.
• The game has no server to measure. ELDEN RING connects players to each other, so the table shows your line to each region and no verdict.
• The region you play on has no test point. The page gives no verdict on it and shows the regions it did measure.
• The region you play on did not answer. No verdict on a region with no figure; run the test again.

What the figure is, and what it is not

1. The test times an HTTPS request over a connection that is already open, to a test point near each region. That is one network round trip, plus well under a millisecond of work at the test point, plus your browser's own overhead.
2. Games send small UDP packets to the publisher's own servers. Those sit on other networks than the test points, often in other buildings or cities, and the route to them can differ.
3. When a packet goes missing, an HTTPS request waits and sends it again, so a loss shows up here as one slow request. A game drops the packet and shows loss instead. Home routers and providers can also treat the two kinds of traffic differently.
4. Games define ping their own way. Some show a pure network round trip, some add server tick or frame time, and publishers do not say which.
5. So the figure is a guide to network distance: it tells tens of milliseconds apart from hundreds, which is what a choice of region or a check of your Wi-Fi turns on. The gap to the game's own number can run either way, because the test point and the game server sit in different places. No honest factor converts one into the other, and the site never shows a game ping estimate.

What your browser adds

Every browser adds a little time of its own to each request. We measured it on October 17, 2026 from our server in a German data center, on a wired line, against curl, a command line tool that times the same requests without a browser. With all test points side by side, as the test runs, Chromium, Firefox and the WebKit engine that Safari uses (its Linux build) added 1.5 to 3.3 ms to the median, and at most 6.3 ms on one test point.

Firefox holds back requests that a page sends back to back. Spacing each test point's requests 0.3 seconds apart removed that delay, which is why the test takes about 10 seconds. A phone browser on Wi-Fi or mobile data has not been measured yet.

Test points that stand in for a city

Every test point is an Amazon Web Services endpoint in the cloud region closest to a game server region. Where a publisher names a city that has no such endpoint, a nearby one stands in, and results on those regions say so:

• North America Central: the test point in Ohio stands in for Chicago, Dallas and Atlanta.
• Middle East (Gulf): the test point in UAE stands in for Bahrain.
• South Africa: the test point in Cape Town stands in for Johannesburg.

AWS reports its UAE region as damaged on its Health Dashboard, so results on the Gulf region carry a warning and a failure there reads as not reachable, never as a figure.

Gaps we show instead of filling

• Latin America South and Turkey (Istanbul) have no test point yet. Their region pages, the map and every game that lists them say no test point yet.
• Russia, the CIS, Manila and Lagos servers that publishers list have no honest stand-in, so the test does not measure them and says so. A Singapore figure would overstate the ping of a player in the Philippines on a Manila server.
• The test points are not game servers (above), and the bands are our own lines.
• Only two statements in the data say which region players from a country land on, and game pages claim nothing beyond those two.
• Some publisher server lists are old, and their game pages print the date of the source.
• VALORANT shards, League of Legends regions, Genshin Impact servers, Free Fire account regions and PUBG MOBILE servers decide where you can play. On those games the test asks which server you play on.
• The only reference measurement on the site is our own server in Germany: one data center line, not a home line, and another network in the same country can read a region tens of milliseconds faster or slower. The region pages print it with its dates. No home line measurement from another country is published yet.

What leaves your browser

The measured times stay in your browser. When a test finishes, the visit statistics record the game, the verdict, the band word, the region names, the country you picked and the settings you ticked, never a millisecond figure. A shared result link carries the figures after the # in the address, a part that browsers do not send to any server. The privacy policy lists the rest.