People are boring and predictable. We have routines that we follow, and even when we break those routines, we do it in the same ways over and over. Using tracking data from volunteers with cellphones and their friends, researchers are able to exploit this to predict where we'll go next, with an accuracy of 60 feet.
If you think about it, your day-to-day schedule is probably predictable without any fancy cellphone tracking tricks. There are set times when you're at home, set times when you're at work, and besides the set times when you're somewhere in between the two plus a few other consistent activities (grocery shopping, the gym, or whatever), nothing much else happens unless your life is far more interesting than the vast majority of people out there. Using just a simple analysis of your past actions, it's fairly easy to predict where you'll be at any given time, except for those rare occasions when you do something CRAZY like go out to dinner. And this is where most predictive tracking algorithms break down: when you throw social events into the mix.
Cellphones, however, are social tools, and researchers from the University of Birmingham took advantage of this to vastly improve predictive tracking accuracy to win Nokia's Mobile Data Challenge. By looking at the location history of one person and combining it with the location history of other people in their address book, this new algorithm can make 24 hour location predictions with an average accuracy of about 20 meters. This is several orders of magnitude better than other algorithms which don't incorporate a social element, and is easily accurate enough to target a SWAT team or drone strike, if you're in to that sort of thing.
Yeah, the obvious worry here is privacy, but if you don't have anything to hide, there's also a lot of potential for predictive location-based services. The researchers suggest a service like Groupon, except since your phone knows where you're headed, it can offer you deals close to where you'll probably be. It would be a compromise, but a compromise that users could make knowingly: an exchange of privacy for better services. Or maybe this is all just a not-so-subtle hint that if your phone can predict your location within 20 meters over a 24 hour period, you really need to get out more.