Epsilon
#basta
Ralph has the team stats on his site, and I've got the goaltenders (Joel Kiers and Frank Currie) on my page - the advanced stats are there, but it's interesting to see how badly save percentage is battered by playing behind a truly-bad team (Currie comes out as 6.1 standard deviations below average ).
Under your binomial assumption, and using the normal distribution to approximate the binomial distribution (appropriate given the number of shots faced), the probability that a goaltender would be that bad is on the order of 0.0000000005, i.e one out of every 2 billion.