Exchange latency from your browser

This test runs in your browser, on your own connection. It opens each venue's public market-data WebSocket and times a request the venue answers, so the number describes the path from where you are sitting — not from our probes.

Nothing is traded and no credentials are used. Only public market-data endpoints the venues document are contacted.

Two numbers, and why they are not the same number

ws_request round trip

Your browser sends the venue's documented ping and waits for its reply. Both timestamps come from your own monotonic clock, so this is a measurement and not an assumption. It is the number to compare venues by.

ws_push_skew clock-dependent

The gap between the timestamp inside a pushed market-data message and the moment your browser received it. This is one-way, and it assumes the venue's clock and your computer's clock agree. If your clock is off, so is this number — it can even come out negative. Shown because it is what other tools report, kept separate because it is not a round trip.

The two are never averaged together and never ranked against each other.

Contributed results are stored without your address: what is kept is the country and network your request appeared to come from, the venue, the method and the timing. Untick the box to keep this run to yourself.

Your results

The test has not been run in this tab yet.

Nothing is contacted until you press Run.

What other visitors measured

Grouped by the country a submission appeared to come from, because an average over every visitor on earth describes nobody. Rows with few samples are exactly that — read the sample count beside the number.

How this test works
What exactly is timed?
For ws_request: a WebSocket is opened to the venue's public endpoint, the venue's own documented ping is sent, and the clock stops when the matching reply arrives. Opening the connection is not counted — the first round trip starts after the socket is open, so TLS and DNS setup do not inflate the figure.
Why can the push figure be negative?
Because it subtracts two clocks that were never synchronised: the venue's and yours. A negative value means your computer's clock is ahead of the venue's. We keep such values rather than hiding them — a number that is quietly clamped to zero looks measured when it is not.
Does a lower number mean faster trading?
No. This measures a network path to a public market-data endpoint. Order entry is a different endpoint, usually a different protocol, and often a different data centre. Nothing here says anything about matching, fills or fees.
Why do the venues here differ from the catalogue?
A venue appears only when a reviewed adapter names its public endpoint and the request it answers. Guessing a protocol from a port number is how a WebSocket endpoint ends up reported as a TCP handshake, so a venue with no reviewed adapter is left out rather than approximated.
What is stored?
Only if you leave the contribute box ticked: the venue, the method, the timing, the outcome, and the country and network your request appeared to come from. Your address is not stored, and no account is needed.