Hacker News Reader

Comment by skc | Hacker News Reader

I've always assumed these people (I see them come across my social media timelines periodically) are just trolling people because I've seen them guess where a place is based on nothing other than a patch of grass in less than 3 seconds.

I still don't see how that is even remotely possible.

View on HN

Almost all of those videos are real. You can in fact quite easily identify most areas down to a few 100s of km just from the surrounding nature. It takes some practice but it is for example very easy to learn which continent the nature is in. The videos you see on social media is also of course the best case out of 100s of games.

Obviously “patch of grass” is going to be harder, but after a while you get good at noticing small details like the style of houses, the way the roads are marked, etc.

I’m by no means a pro but I went through a hyperfixation on Geoguessr YouTube videos so I started picking up the general architectural styles of Nordic houses for instance. Or the specific markings you can only find in Russia.

There’s also image artifacts which are only present in specific collection attempts. Most images from IIRC Ghana have a specific artifact and can easily narrow it down. There’s a few of these out there.

If the site is just using google api in the html source it's probably not too hard to extract the location from the javascript. A quick search shows there are browser scripts for that. But they could protect against by calling the streetview api from their server (if google allows that) and only serving the images to the user.

I play quite a bit, although I'm much more casual and generally just focus on the US. I find myself identifying areas based on foliage pretty frequently. They can be super localized and after you play a few hundred or thousand maps, you just start naturally associating different areas with different plants. I'm nowhere near good enough to guess in 3 seconds based off anything, but after a few months I find I can generally spawn, not move the camera, and correctly guess I'm in 1 of like 4 states. Depending on what's visible, I can often identify the correct state immediately, although pinpointing takes a lot more time. It's just like any game: repetition and practice makes perfect.

Reply on news.ycombinator.com