Restrict yourself to explanations which can be rendered as a series of individually simple steps--something you could reduce to math; something you could program into a computer. Note that "magic" does not qualify, here.
The shortest such series of steps that successfully outputs your observation is the most likely explanation for those observations. In this case, we would either need an exception to well-known rules of fluid dynamics, or an explanation why 2 years of searching in a small area yielded no meteorite. Either one of those is an added complication to the explanation which predicts stowaway gravel: P(A & B) <= P(A).
The shortest such series of steps that successfully outputs your observation is the most likely explanation for those observations. In this case, we would either need an exception to well-known rules of fluid dynamics, or an explanation why 2 years of searching in a small area yielded no meteorite. Either one of those is an added complication to the explanation which predicts stowaway gravel: P(A & B) <= P(A).