Physics being what it is only your lowest ping will be relevant.

...
Closing in

The new method zooms in through three stages to locate a target computer. The first stage measures the time it takes to send a data packet to the target and converts it into a distance � a common geolocation technique that narrows the target's possible location to a radius of around 200 kilometres.

Wang and colleagues then send data packets to the known Google Maps landmark servers in this large area to find which routers they pass through. When a landmark machine and the target computer have shared a router, the researchers can compare how long a packet takes to reach each machine from the router; converted into an estimate of distance, this time difference narrows the search down further. "We shrink the size of the area where the target potentially is," explains Wang.

Finally, they repeat the landmark search at this more fine-grained level: comparing delay times once more, they establish which landmark server is closest to the target. The result can never be entirely accurate, but it's much better than trying to determine a location by converting the initial delay into a distance or the next best IP-based method. On average their method gets to within 690 metres of the target and can be as close as 100 metres � good enough to identify the target computer's location to within a few streets.

Client independent

That kind of accuracy normally requires people to deliberately disclose their location, but Wang's method works without the user's permission. "This is a client-independent method," as he puts it. "The client does not need to approve anything."
...

In reply to a post by john2007:

Physics being what it is only your lowest ping will be relevant.

I wasn't thinking of physics, I was thinking of BT's MSANs

I only know one person who works less than 1/2 mile away from his home, so I guess my expectations are coloured by that.

The census will let us know.

Blocking ICMP echo requests won't win you much at all. You could solicit the exact same information using the initial stages of a TCP 3-way handshake, or even just a SYN+RST. Plus if you wanted you could collect the data in the opposite direction (i.e. have JS make and time the round trip of a request to an echo server) and report that data back.

Wouldn't bog standard NAT reject unsolicited requests?

I think it's probably one of those techniques that works well in theory, and passably well on initial tests with well-defined and co-operating systems.

But when it gets out into the real world... we've all seen people on here wanting to know why they get 80ms pings to their favourite game server and their next-door neighbour gets 15ms

For reject I should have said ignore.

Exactly. And so you measure the time from the unsolicited request to the receipt of the rejection. Round trip time measured with no ICMP echo request needed.

Not quite the point. Final leg pings (exchange to home) should be consistent.

My router ignores unsolicited requests.