# Yao's Minimax Principle with error

In Yao's paper proving Yao's principle, he equates the lower bound of a randomized Las Vegas algorithm with the expected run time of the best deterministic algorithm on the worst input. He goes on to generalize this to Monte Carlo algorithms with error, in particular in Theorem 3:

However I do not understand how this result follows from the original principle. Could someone please explain/shed some color?

Cheers