14 out of 32 first rounders were three star recruits with two being two star. link
In that case the high probability 5 and 4 stars are vastly outnumbered by the lesser probability 3 stars, so the odds may work out in the 3 stars favor.
Or think of it this way: 30 of ~2500 recruits will be the best after 4 years. The odds of a 5 star being one is 50%, 4 star 25%, 3 star 10%. 100 5 stars, 500 4 stars, 10000 3 stars (guessing here). That comes out to 50, 125, and 1000, so the odds of one being a 3 star is much higher, 1000/1175. It sounds wrong, but that's how Bayesian probality works. For example, for a medical test that's 99% accurate, for a disease that has an incidence in the population of 1%, the odds that a person that tests positive has the disease is 50%. Take a sample of 10000 people: 100 with disease, 9900 without, due to 1% incidence. For 100 with, 99 will test positive (99% accuracy). For the 9900 without, 99 will test false positive (99% accuracy). So 50% odds for anyone who tests as positive (50% true positive, 50% false positive).
