Packet Loss Statistics From Real Tests
Read this before quoting any number below. These are measurements taken by people who came to a packet loss test — which is not a random sample of anything. Most arrive because something is already wrong. Every figure here therefore describes the tests run on this site, and none of them describes a country's internet. "42 % of tests from country X saw packet loss" is a true sentence about this sample; "42 % of connections in country X have packet loss" is not, and would be wrong by a wide margin.
With that said, the sample is large, it is real UDP rather than a TCP approximation, and it is collected the same way everywhere — so it is useful for comparing like with like: the same kind of visitor, with the same kind of problem, measured with the same method.
Across all 106,813 tests
Two things drive that first figure, and both need saying before anybody quotes it. The obvious one is the sample: almost everyone here came because something was already wrong. The less obvious one is arithmetic — a test counts as "saw loss" if a single packet failed to return, and a two-minute test at 128 packets per second sends more than fifteen thousand of them. The longer and faster the test, the more likely it is to catch one. That is the correct behaviour for a diagnostic tool, and it makes this column a poor measure of severity. The next column along, tests above 1 % loss, is the one that tracks "bad enough to feel".
By country of the person testing
Countries with fewer than 100 stored tests are left out entirely rather than shown with a number that two unlucky measurements could move. Ordered by how many tests there are, not by how bad they look.
| Country | Tests | Tests that saw loss | Median loss, when there was loss | Median ping | Median jitter |
|---|---|---|---|---|---|
| Russia | 25,985 | 76.83 % | 14.78 % | 71 ms | 5 ms |
| United States | 9,297 | 65.86 % | 1.14 % | 43 ms | 5 ms |
| Germany | 7,796 | 69.28 % | 6.01 % | 48 ms | 8 ms |
| Indonesia | 5,622 | 74.23 % | 2.71 % | 39 ms | 7 ms |
| India | 3,386 | 75.58 % | 1.29 % | 88 ms | 6 ms |
| Iran | 3,276 | 73.63 % | 20.81 % | 130 ms | 16 ms |
| Netherlands | 2,753 | 72.36 % | 8.97 % | 86 ms | 9 ms |
| Turkey | 2,719 | 69.66 % | 3.9 % | 70 ms | 7 ms |
| Ukraine | 2,544 | 72.01 % | 2.46 % | 55 ms | 5 ms |
| Canada | 2,506 | 66.12 % | 1.27 % | 46 ms | 6 ms |
| Poland | 2,282 | 65.29 % | 1.86 % | 47 ms | 7 ms |
| Philippines | 2,279 | 75.38 % | 2.57 % | 72 ms | 7 ms |
| Egypt | 2,152 | 55.95 % | 4.31 % | 81 ms | 9 ms |
| United Kingdom | 1,922 | 59.73 % | 2.85 % | 42 ms | 7 ms |
| Finland | 1,312 | 72.79 % | 8.55 % | 79 ms | 8 ms |
| France | 1,286 | 68.74 % | 4.97 % | 33 ms | 8 ms |
| Sweden | 1,277 | 69.15 % | 1.57 % | 74 ms | 7 ms |
| Algeria | 1,199 | 80.65 % | 4.52 % | 73 ms | 8 ms |
| Italy | 1,105 | 51.22 % | 3.38 % | 58 ms | 8 ms |
| Bangladesh | 1,073 | 77.91 % | 2.1 % | 64 ms | 8 ms |
| Colombia | 1,028 | 84.05 % | 1.96 % | 100 ms | 6 ms |
| Georgia | 915 | 70.05 % | 0.02 % | 66 ms | 2 ms |
| Brazil | 868 | 62.33 % | 1.82 % | 142 ms | 4 ms |
| Saudi Arabia | 805 | 53.79 % | 1.81 % | 103 ms | 8 ms |
| Kazakhstan | 697 | 67.58 % | 2.06 % | 204 ms | 4 ms |
| Malaysia | 674 | 64.39 % | 2.3 % | 47 ms | 10 ms |
| Australia | 673 | 60.92 % | 1.02 % | 159 ms | 4 ms |
| Belarus | 657 | 56.93 % | 1.96 % | 57 ms | 4 ms |
| Iraq | 648 | 55.25 % | 4.73 % | 90 ms | 10 ms |
| Vietnam | 618 | 71.36 % | 1.71 % | 60 ms | 4 ms |
| Pakistan | 607 | 69.85 % | 3.49 % | 138 ms | 7 ms |
| Japan | 573 | 65.62 % | 3.3 % | 109 ms | 5 ms |
| Switzerland | 517 | 71.95 % | 6.3 % | 78 ms | 11 ms |
| Mexico | 515 | 57.48 % | 3.06 % | 81 ms | 4 ms |
| Venezuela | 512 | 57.03 % | 1.82 % | 88 ms | 2 ms |
| Hungary | 503 | 68.99 % | 1.65 % | 35 ms | 8 ms |
| Czechia | 500 | 69.6 % | 1.19 % | 29 ms | 7 ms |
| Israel | 483 | 53.42 % | 0.62 % | 66 ms | 3 ms |
| Singapore | 481 | 70.27 % | 8.36 % | 66 ms | 7 ms |
| Norway | 471 | 65.61 % | 2.57 % | 48 ms | 8 ms |
| Spain | 463 | 64.58 % | 2.57 % | 62 ms | 7 ms |
| South Africa | 451 | 67.63 % | 1.57 % | 208 ms | 10 ms |
| Austria | 450 | 66.89 % | 1.84 % | 45 ms | 9 ms |
| Latvia | 425 | 72.0 % | 8.22 % | 87 ms | 10 ms |
| Hong Kong | 396 | 68.18 % | 5.62 % | 101 ms | 6 ms |
| Portugal | 391 | 69.57 % | 0.64 % | 66 ms | 6 ms |
| Slovakia | 370 | 70.27 % | 0.04 % | 55 ms | 4 ms |
| Estonia | 359 | 64.35 % | 7.55 % | 86 ms | 10 ms |
| Peru | 357 | 59.1 % | 0.66 % | 115 ms | 3 ms |
| Thailand | 354 | 68.64 % | 0.78 % | 44 ms | 5 ms |
| China | 347 | 84.15 % | 14.1 % | 256 ms | 10 ms |
| Belgium | 331 | 55.29 % | 1.19 % | 42 ms | 9 ms |
| Greece | 315 | 61.9 % | 6.08 % | 59 ms | 8 ms |
| Denmark | 309 | 64.4 % | 2.55 % | 48 ms | 7 ms |
| Morocco | 305 | 61.64 % | 2.0 % | 66 ms | 9 ms |
| Argentina | 287 | 59.93 % | 1.44 % | 173 ms | 5 ms |
| Bulgaria | 282 | 73.76 % | 6.11 % | 100 ms | 10 ms |
| Uzbekistan | 236 | 63.56 % | 1.23 % | 242 ms | 4 ms |
| Romania | 229 | 54.59 % | 3.85 % | 44 ms | 5 ms |
| Lithuania | 225 | 58.67 % | 11.78 % | 73 ms | 6 ms |
| Tunisia | 221 | 76.47 % | 5.68 % | 72 ms | 13 ms |
| United Arab Emirates | 214 | 60.28 % | 7.16 % | 178 ms | 9 ms |
| LY | 213 | 66.67 % | 8.93 % | 89 ms | 13 ms |
| Ecuador | 204 | 48.04 % | 0.64 % | 105 ms | 3 ms |
| Croatia | 185 | 29.73 % | 0.55 % | 36 ms | 6 ms |
| Taiwan | 180 | 81.11 % | 3.92 % | 89 ms | 5 ms |
| Lebanon | 176 | 82.39 % | 4.28 % | 92 ms | 11 ms |
| Nepal | 164 | 69.51 % | 1.33 % | 99 ms | 4 ms |
| Serbia | 154 | 59.74 % | 1.4 % | 37 ms | 7 ms |
| Cambodia | 140 | 88.57 % | 3.44 % | 77 ms | 13 ms |
| Moldova | 135 | 58.52 % | 3.41 % | 77 ms | 12 ms |
| Ireland | 132 | 64.39 % | 3.09 % | 61 ms | 12 ms |
| Azerbaijan | 129 | 67.44 % | 11.32 % | 114 ms | 12 ms |
| Chile | 128 | 73.44 % | 1.15 % | 157 ms | 6 ms |
| Kenya | 117 | 72.65 % | 5.7 % | 181 ms | 13 ms |
| North Macedonia | 116 | 82.76 % | 4.08 % | 51 ms | 6 ms |
| New Zealand | 110 | 53.64 % | 1.05 % | 186 ms | 5 ms |
| SY | 108 | 89.81 % | 22.56 % | 105 ms | 16 ms |
Medians, not averages: a single half-hour test on a broken line would otherwise drag a whole country with it. The loss column counts tests, not packets — a test appears in it if even one packet failed to come back within a second.
By measurement server
The same tests grouped by which of the four servers they were measured against. This one says more about distance than about anybody's line: a visitor measuring from Jakarta to Frankfurt is measuring half the planet, and the ping column shows it.
| Server | Tests | Tests that saw loss | Median ping | Median jitter |
|---|---|---|---|---|
| Europe (Frankfurt) | 66,409 | 71.31 % | 68 ms | 7 ms |
| Asia (Singapore) | 21,377 | 71.73 % | 84 ms | 6 ms |
| US East (New York) | 13,870 | 66.61 % | 74 ms | 5 ms |
| US West (San Jose) | 5,157 | 61.99 % | 60 ms | 5 ms |
How these numbers were produced
- Every row is a finished test of at least 10 seconds with a real measurement in it. Failed connection attempts, which record a 0 ms ping, are excluded.
- The window is the last 365 days, so a routing change from two years ago no longer shapes what is claimed today.
- Location is the country of the person running the test, derived from their IP address at the time of the test. The IP address itself is never stored — see the privacy policy. Cities are not published: a city row would be a small enough group to start describing identifiable people, and it adds nothing a country row does not.
- No ISP or network operator is named anywhere on this page, and that is deliberate: the same sampling bias applies with much sharper consequences when a company is named, and "provider X has 4 % packet loss" is a claim this data cannot carry. Providers with enough stored tests to show a distribution — which is a different statement, and the only one this sample supports — have their own pages under network providers, each carrying that caveat at the top.
- Measured over real UDP through WebRTC DataChannels, at the packet rate of the game the visitor picked. The full method is on the methodology page, including what it cannot determine.
- The measurement itself changed in 2026. Before that correction, the reported average ping was biased low for tests at high tick rates — the chart kept the fastest packet of each group instead of a representative one. Rows stored before the fix carry that bias, so the ping columns here should be read as a rough level, not to the millisecond, until a full window of tests has been measured the new way. The share-of-tests-with-loss columns are not affected: they were always counted per packet.
- Rebuilt once a day. This table was generated on 2026-09-19.
Using these figures
You are welcome to quote or reproduce anything on this page, in an article, a forum post or a complaint, with a link back to this page. One request: keep the qualifier attached. The correct form is "of tests run on PingPacketTest.com", not "of connections in" — and if you find a figure here that does not hold up, write to me and it gets corrected.
Want to add a measurement of your own? The test is running by the time the page has loaded, and every finished result can be exported as a PDF report for your provider.