Inspired by the first series of the BBC's excellent Sherlock, I once ran a small blog in which I would take photographs of random people I saw on the tube/bus/train and upload them along with any deductions I could make. It was hardly a big hit - in the three months it was active it made just enough money for me to buy the box-set of Sherlock Series 1, and then I had exams the like and took it down (oh, there were also a few... complaints). However, I never really got out of the habit of attempting deductions, which, sadly, are rarely accurate. In my defence, it's much harder than Sherlock makes it look.
An example: this morning, walking home from the supermarket, I passed a man standing at the corner of a building talking on his phone. Some things were immediately obvious - his suit and shoes showed an indoor office worker; he was married, neat and careful of his appearance. It was pretty likely that he worked in the bank branch against which he was leaning. It was his phone, however, that caught my attention: it was an iPhone 4S, with a yellowing plastic-rubber cover. This discolouration of the cover started a line of reasoning: it looks old, definitely older than the phone, which could only be four months old at the most (release date was in October). One could therefore assume that he had previously owned the iPhone 4, and combining that with the brand loyalty that Apple products tend to inspire, one could assume that he was a long-time Apple user.
It was a small but interesting deduction, and I was feeling rather pleased with myself before I rounded the corner and realised the obvious flaw in my logic. The case was white. The iPhone 4 was black. A man who took the time to shave his rather odious goatee every morning and polished his shoes so regularly would almost certainly not put a white case on a black phone, and thus the case was either bought secondhand (extremely unlikely), or it was just an unpleasant colour from the get-go.