That sounds like someone told him that about a third wouldn't receive a 5* BH and he transcribed it as "about 33%", which implies a higher accuracy.
It is pretty easy to calculate the probability of not receiving a bonus hero in 167though:
>>> p = 0.15 * 0.05 >>> (1 - p) ** 167 0.2844435417820726
To keep that in perspective, we can also calculate the probability of receiving exactly one 5* BH in 167 pulls:
>>> 167 * p * (1 - p) ** 166 0.35895771897435347
That means the remaining 36% would receive multiple 5* BH covers. In other words, more people would receive 2+ covers over 167 pulls than the number of people who received 0 covers.