Certainly a reasonable point, but a bit shortsighted. What you are saying, essentially, is that momentum invalidates one key assumption of the binomial distribution, specifically the assumption that successive trials/racks are statistically independent. Statistics, as you seem to understand, only measure what will happen on average, not in this very moment.
Of course, most well-known probability theory is based on the assumption that successive trials are statistically independent, even though the assumption is often not quite right. Fargo Rate works this way, basing your rating and your implied win probability against an opponent on average performance, ignoring qualitative factors like rivalry, momentum, etc.
So no, binomial distribution methods are not out the window at all here, as they represent the best estimates available given well-known statistical methods, but equally true is that the limitations of using probability theory in this, or any other, way need to be understood.