Latency
Updated 2026-10-05T19:52:10Z
Measured continuously since 4 Oct 2026: 365,366 updates, median 2.5 ms, 95th percentile 12 ms, 99th percentile 33 ms, from a price change reaching our servers (Helsinki) to the update arriving at a client.
| Period | Updates | Median | p95 | p99 | p99.9 |
|---|---|---|---|---|---|
| Last minute | 93 | 1.6 ms | 5.9 ms | 10 ms | – |
| Last 7 days | 365,366 | 2.5 ms | 12 ms | 33 ms | 183 ms |
| Last 30 days | 365,366 | 2.5 ms | 12 ms | 33 ms | 183 ms |
| Since 4 Oct 2026 | 365,366 | 2.5 ms | 12 ms | 33 ms | 183 ms |
What we measure
The time from a price change reaching our servers to the update arriving at a client. These are the same two points other Pinnacle feeds publish as "frame to client". Every update counts, not a sample, and the figures are recomputed every minute. The same numbers are at latency.json.
What it doesn't include: the time between Pinnacle changing a price and Pinnacle publishing it. That happens inside Pinnacle's systems, so no provider can measure it from outside.
Your network
Our servers are in Helsinki, Finland. Your own network time comes on top of the figures above. Estimated one-way time to a client in:
| Country | Add |
|---|---|
| France | +21 ms |
| Netherlands | +16 ms |
| Romania | +28 ms |
| Italy | +20 ms |
Measured every 15 minutes through servers in each country: a quarter of the median round trip there and back through them. An estimate, averaged over the last 7 days.
To measure your own: every stream message carries sentAt, the time our server sent it. The docs have a 10-line script that prints your median and p95.