This article makes no damn sense. Not only does it fail to explain the nonsensical-sounding theories which they used to predict it, it also says:
According to a probabilistic model they created, there was an 80.9% chance that bin Laden was hiding out in Abbottabad, Pakistan
Wow, really? But...
To be clear: the class identified the nearby city of Parachinar as being the most likely hideout.
So there was an 80% chance he was in Abbotabad, but they considered it more likely that he was in Parachinar?
The whole thing sounds about as useful as the psychics who will no doubt come out of the woodwork over the next few days to say that they too correctly predicted UBL's hiding spot.
It's a bastardized summary of a technical paper; this is not a surprise if you've ever read _USA Today_ trying to explain what XML is.
That said, they link the paper so you can read it for yourself. So far it makes a lot of sense to me; turns out you can eliminate a great many buildings and locations by selecting a few characteristics.
For example, bin Laden's height, need for dialysis, entourage, and defensible walls rules out all but 16 structures in the Kurram region (http://en.wikipedia.org/wiki/Kurram_Agency) if you look for only 5 out of the 6 criteria. (Page 4 of PDF.)
As for your specific criticism; you're quite right. The OP has misstated his link, which says:
> According to a probabilistic model they created, there was an 88.9% chance that bin Laden was hiding out in a city less than 300 km from his last known location in Tora Bora: a region that included Abbottabad, Pakistan, where he was killed last night.
That's much more sensible, and does follow directly from their geographical theory - if you think of animals as random particles moving in Brownian motion, there is an exponential dropoff. Random walks don't go very far.
On a side-note, they were largely right about the room characteristics. The compound bin Laden was in was in fact, a tall building with multiple structures with more than 3 rooms and walls over 3 meters tall (http://en.wikipedia.org/wiki/File:Osama_bin_Laden_hideout.jp...).
In fact, the only characteristic I see missing is 'trees for cover when outside' (few trees in http://en.wikipedia.org/wiki/File:Osama_bin_Laden_hideout.jp... ) and arguably, the need for aerial cover was met by being so near the military academy. (I've seen reports and speculation that the helicopters had to spoof Pakistani transponders to avoid triggering air defenses.)
Y'see, that's much more sensible... an 89% chance that he was within 300km of his last known location, given that he can't exactly get on a plane, sounds way more plausible than an 89% chance that he was in one particular medium-sized Pakistani town.
(Of course it's also unfalsifiable to say that he has a 90% chance of being somewhere...)
And yet the nature of the internet news cycle is such that I'll be reading for weeks to come that a group of geographers correctly predicted two years ago that he was in Abbottabad and that anyone with google maps could have found him ages ago. Sigh.
I think the original article from ScienceInsider realized they had it wrong and have corrected it...
"According to a probabilistic model they created, there was an 88.9% chance that bin Laden was hiding out in a city less than 300 km from his last known location in Tora Bora: a region that included Abbottabad, Pakistan, where he was killed last night. "
It looks like a nice little research project that didn't take itself too seriously, but the article reads to me like "he's probably a few tens of km inside Pakistan because it's not too far from where he was last seen 10 years ago, go look in these three buildings".
Of course, the prevailing view was that he was somewhere in the border region and he turned out not to be in any of those buildings.
An excellent exploration of the art of understatement:
The derivation of the model also seems rather ad hoc in nature (fitting a power law to the assumption of a 99% intensity at Tora Bora and a 1% intensity at Washington DC).
Yep, that sounds pretty arbitrary! (I'd also be amazed if it was only 99 times less likely for him to be in Washington DC than Tora Bora...)
According to a probabilistic model they created, there was an 80.9% chance that bin Laden was hiding out in Abbottabad, Pakistan
Wow, really? But...
To be clear: the class identified the nearby city of Parachinar as being the most likely hideout.
So there was an 80% chance he was in Abbotabad, but they considered it more likely that he was in Parachinar?
The whole thing sounds about as useful as the psychics who will no doubt come out of the woodwork over the next few days to say that they too correctly predicted UBL's hiding spot.