In fact, the inequality in Bell's theorem is based exactly on how classical statistics works: if you and I randomly choose to measure some aspect of each of a pair of "entangled" objects, and assuming the result of our measurement can only be +1 or -1, then on average the sum of our measurements will be less than or equal to 2. It turns out though that this logic doesn't work for entangled quantum objects.

And one small note here: based on everything we know, the key here is quantum entanglement, not scale. That is, if you could entangle two basketballs or planets for long enough to perform a Bell test on them, they would likely reproduce the particle results. However, this property of quantum systems is very hard to preserve for such a large system with so many ways of interacting with the environment and experiencing decoherence.