"On the other hand, if no restrictions whatever are imposed on the hidden variables, or on the dispersion free states, it is trivially clear that such schemes can be found to account for any experimental results whatever. Ad hoc schemes of this kind are devised every day when experimental physicists, to optimize the design of their equipment, simulate the expected results by deterministic computer programs drawn on a table of random numbers." (1)
I found such a limit mentioned in the arxiv article from Reid et al. in which they summarize the current (well..., 2008) state of the EPR debate:
"The original Bell inequalities requires a threshold efficiency of 83 % (η ~ 0.83) per detector (Garg and Mermin (1987); Clauser and Shimony (1978); Fry et al. (1995)), in order to exclude all local hidden variable theories. For lower efficiencies, one can construct local hidden variabe theories to explain the observed correlations (Clauser and Horne (1974); Larsson (1999))." (2)
- Introduction to the hidden variable question, J.S.Bell, 1970, http://cdsweb.cern.ch/record/400330/files/CM-P00058691.pdf
- The Einstein-Podolsky-Rosen paradox: from concepts to applications, M.D. Reid et al.2008, http://arxiv.org/PS_cache/arxiv/pdf/0806/0806.0270v2.pdf
- Using linear programming to construct better criteria for closing the detection loophole in EPR experiments, James H. Bigelow, 2008, http://arxiv.org/ftp/arxiv/papers/0805/0805.0387.pdf