Perhaps the 20% strike rate was in reference to races in which number 1 was not scratched? In that sample of 103,000 where #1 won 16% of them, perhaps the remaining 4% were races from which it was scratched?