His math is suspect, but he has the right idea. While the percentage chance of success never changes the probability of continued failure does go down. On the first attempt the odds of success versus failure are 20%/80%, however the odds of success versus seven failures in a row is all but a coin flip - 20% chance of success versus 20.97% chance of seven failures. And the odds of success are actually better than that of eight failures 20%/16.78%
There no such thing as "the probability of continued failure". There's only (a) the probability of a single attempt, and (b) the probability of learning m
schems in n
attempts. Assuming we consider m
>= 1, then the probability mentioned in (b) is higher for high n
than low n
. But it doesn't matter at all whether you have previously failed; no probabilities change when you continue to fail (or succeed).