In many worlds, all the results with probability>0 happens. But the predictive power of a theory is based in that events with very low probabilities no matter ( the possible ocurrence is `[but a constant]) around the averaged valued by the exponential of shannon entropy, that in the case of gaussians distributions, it is the typical standard deviation) Without this limitation, we could flip a coin 100 times and obtain 100 faces, probability is not 0 is 1/2^(100). What is the Everet III solve of this problem? I read his thesis but not mention to it. HOw is this trouble solved??