lemme help wrote:
There is a weakness in the current "statistical argument" against having only 1 photo. It is always possible for a photographer to take a break, check his phone, etc and miss some runners. If this happens, then it's not just 1 photo lost, but 2-3 photos lost (most runners have several photos at each site). So Rossi can reasonably claim his 1 photo is a result of bad luck at several locations. After all, in a large enough sample, outliers *do* occur.
I already responded to this once to point out that what we did statistics on was the number of locations where runners were captured, not the number of photos. As a result this criticism misses the mark, and our stats remain valid. However, just for the sake of intellectual curiosity I wanted to point out that even if we had done stats on number of photos, the greater numbers of photos being lost at each location would simply translate into a higher value of the standard deviation for that data set. As a result, the fact that Mike Rossi would be further from the average would not translate into a spuriously low probability as you seemed to fear. (Actually, add in the fact that there are like 5 pictures of Mike crossing the finish line, and if anything this method wouldn't condemn Mike as badly.)
But anyway, the real problem with the # of photos method would be that the distribution probably wouldn't be as normal, because smaller numbers of factors (getting missed at a smaller number of locations where more pictures were taken) might have a bigger impact. This would make the z-test method less valid.