I'm a casual horse racing fan and I think several reasonable contributing hypotheses have been aired in this thread. Let me toss mine in. It seems to me that the way thoroughbred race horses are trained is extremely unscientific. This seems kind of nuts considering the sums of money involved, and since I'm only a casual fan perhaps I don't know what I'm talking about. But whenever I hear interviews with horse trainers, even the most successful ones, it seems that they decide how to work out and race their charges based largely on "seat of the pants" intuition and what they've always done before, not on any quantitative analysis.
Taking the data seriously has really changed many sports including baseball and football. I don't know why the same approach isn't being applied to horse racing, which seems hopelessly mired in superstition and tradition.