Most substances can crystallise into two or more different crystal lattices, called polymorphs. Despite this, there are no systems in which we can quantitatively predict the probability of one competing polymorph forming, instead of the other. We address this problem using large scale (hundreds of events) studies of the competing nucleation of the alpha and gamma polymorphs of glycine. In situ Raman spectroscopy is used to identify the polymorph of each crystal. We find that the nucleation kinetics of the two polymorphs is very different. Nucleation of the alpha polymorph starts off slowly but accelerates, while nucleation of the gamma polymorph starts off fast but then slows. We exploit this difference to increase the purity with which we obtain the gamma polymorph by a factor of ten. The statistics of the nucleation of crystals is analogous to that of human mortality, and using a result from medical statistics we show that conventional nucleation data can say nothing about what, if any, are the correlations between competing nucleation processes. Thus we can show that, with data of our form, it is impossible to disentangle the competing nucleation processes. We also find that the growth rate and the shape of a crystal depends on when it nucleated. This is new evidence that nucleation and growth are linked.