I'm only interested in S17.My EV for doubling is only slightly higher than yours probably explicable by my infinite deck approach versus your fixed subset approach.But my figure for splitting is more than 3% lower than yours which means that one of us must have in error in the program.I assume it's me because there was already a difference with figures quoted by MathProf. I will tell you as soon as I have found the bug.
Francis Salmon