My simple log-linear model forecasts a winning time of 9.68 seconds, which is 1/100 of a second faster than Usain Bolt's winning time in Beijing in 2008, but still 1/10 of a second slower than his 2009 World Record (9.58s) in Berlin.
Markus used the readHTMLtable function to scrape the data from the web into R, and the lm function to fit a regression model and make the prediction (the red line above). (The R code is provided at Markus's blog.) He admits the model is overly simplistic: it doesn't use data from non-Olympic competitions, and a model like this implies that 100m sprint times will shrink indefinitely. But you can see how well this prediction pans out this Saturday, in the men's 100m final. (Update: The winning time was actually 9.63 seconds.)
mages' blog: London Olympics and a prediction for the 100m final