I’m not sure I can fit the dice analogy with actual testing methods. The variables are more complex and due to practical limitations, Alice isn’t actually able to positively identify the numbers that appear on the dice.
I think it would be better to describe an actual experiment.
Alice and Bob are at opposite ends of an optical table. Alice and Bob each have a polarizing filter. In the middle of the table, an emitter sends out a photon pair with identical polarity. If Alice and Bob’s filters are both vertical, then Alice and Bob should see the same number of photons pass through their filters and strike their detectors (within the limits of the detectors efficiency, which is statistically non-trivial with today’s technology). If their filters are perpendicular, then Alice and Bob’s results should oppose each other.
Parallel…
Bob: I see it!
Alice: So do I!
Perpendicular…
Bob: I didn’t see it.
Alice: I did.
If the two filters are at some other angle, the percentage of agreement should be roughly the squared cosine of the angle between the two detectors. eg. 45 degrees between the filter axes, should result in 50% agreement.
In experiments, the photons do follow this scheme (again within the limitations of the detection apparatus). But, when several runs of the experiment are performed, with the angles altered between runs and the results are all composed into a single probability distribution, the percentages seem to be breaking with Bell’s Inequality and by extension, must either be breaking quantum rules of ‘locality’ or ‘reality’.
The problem is, that it is incorrect to assume a single probability distribution for multiple, different configurations of the apparatus. This “Joint Measurability Assumption” (as Thomas Brody calls it) is like Alice and Bob saying "If we had left our polarizers in the 1st configuration, when whe made the 2nd run, the results would have been the same as the first run. Since there is no sound basis for that assumption, the resulting conclusion is meaningless (especially in the case of looking for small statistical anomalies in a completely random random event).
I hope that made sense, this little composition window makes proofing rough.
Stephen
Stephen’s Website
Satellite Hunting 1.1.0 visible satellite pass prediction
shareware available for download at
Satellite Hunting