To rephrase: there is no evidence to suggest it matters
Also don't overlook the random distribution. If the odds of producing a good horse from a mating between "poor sire" and "poor mare" (loosely defined) is 1 in 500 (I am just randomly picking a number to illustrate the point so don't get hung up on it), then 1 in every 500 will be good and the reason is just the variation involved, i.e. produce enough foals and *at some point* there will be good ones just by virtue of the number of foals. There is no further explanation to it and no more logic or explanation to be read from it