As Stacy McGaugh likes to put it in his blog - the question is why MOND (a relatively simple theory with just one free parameter) is able to make successful predictions at all? Predictions like this one, which you have no reason to expect to be true in CDM models?