FrediFizzx wrote:gill1109 wrote: So it's not clear to me, Fred, whether you are against any kind of CHSH type experiment in principle, which seemed to be Han Geurdes' position, of if you simply have a different CHSH protocol in mind from what other people have used in the past.
I am not against CHSH for a Weihs, et al, type of experiment. That is what CHSH was designed for. And the exact protocol for CHSH is,
E(a, b) + E(a', b) + E(a, b') - E(a', b')
Where the minus sign can be transposed to any one of the other elements and it is usual practice to take the set that gives the highest absolute value. Now, you seem to have a hard time understanding exactly what that protocol means as it
does not mean <AB> + <A'B> + <AB'> - <A'B'>. You are mixing up results with angle settings.
However, CHSH is totally un-necessary for Joy's experiment. All that Joy's experiment has to show is E(a, b) =
-a.b as Han is saying.
Obviously if E(a, b) =
-a.b then E(a, b) + E(a', b) + E(a, b') - E(a', b') exceeds 2 for skillfully chosen a, a', b, b'. Violating CHSH for
particular a, a', b, b' is *weaker* than showing E(a, b) =
-a.b for all a, b. If Joy does what Han wants then he even gets CHSH = 2 sqrt 2, where I will be defeated if CHSH is appreciably larger than 2.
I have no idea what you mean by the distinction between <AB> and E(a, b). If we perform, say, n runs with settings a, b on each side of the experiment, then we will compute the average of the products of the outcomes on each side of the experiment. I used the notation <AB>_obs to stand for this quantity. That's what experimenters talk about, that's what they make plots of. That's what Joy and my bet is about.
Joy writes in his theoretical papers about the limit for n to infinity of the same quantity, but n is not going to go to infinity in the experiment - we will have to make do with some finite, large, n. The bet has to be settled in finite time and with finite resources. Physicist's don't observe mean values. They observe averages and they know that they are close to means, because of the law of large numbers. They even draw error bars in the graphics in their experimental papers. Joy and I will choose n large enough that the error bars will be so small that it is clear who has won.