I've seen this before, and it's cool, but I think it's better for prediction of shorter race times (under the 1/2 M). If I use it to try to predict my current marathon time, it's somewhere between 3:39-4:05, depending on what race time from the last 5 months I use (In 5 months I've done everywhere from a 1 miler to a 1/2 M). That's a huge window! I'm not a good short-distance runner, and typically place better in the longer races, so I know the times from my longer races (10 miler, 1/2 M) are probably a better indicator of how I'll do. I really think for a marathon, it's easier to predict a time based on training and a better indicator would be how one does for in tempo runs and MP runs. I think anyone using it definitely has to consider other factors and know it's not an exact science. (Just my unprofessional opinion!)