For being 15 year old, cool work!
But calling this de-anonymization is a stretch, if it can possibly pinpoint you within 250 miles (that's assuming geoip is correct too, which it rarely is).
In their GeoGuesser demonstration video, the higlighted area is densely populated and you still would need to match millions of people vs the online user.
It does provide some hints as to the location of the targeted user, and that is cool!
It is already more than enough to know which country to contact the authorities and to pinpoint a rough area where to look.
If the scammer is in Nigeria, tough luck. If he is in the EU or US then exists a feasible chance to go after the person.
> assuming geoip is correct too
It's not using geoip, it's using anycast.
De-anonymization would take monitoring over a period of time, but it could definitely work. Take this scenario for example: a person of interest is in the area of New York on Jan 1. On Jan 4 they travel to the UK. On Jan 7 they travel to Germany. On Jan 21 they travel back to the US.
The list of suspects would be fairly small when US officials cross-check individuals that travelled US-UK on Jan 4 and Germany-US on Jan 21.