In the usual experiment used, if the two electron spin detectors for the two entagled particles are tilted the same, they'll get opposite results always. But if you tilt them differently, however, you no longer get results that are exactly opposite each other all the time. Therefore the electrons are in a superposition until they hit the detectors. It isn't decided yet what they will be.
Keep in mind we're working with quantized values. Spin is always completely up or down, no matter what angle you measure it from. Measuring it at that angle locks it down, but tilting another detector in the chain after means you go back to a non-linear probability calculated by quantum mechanics (cosine of the angle in this case) to if it is up or down again at the new tilt and locks it down to that new tilt. There isn't any pre-determined value the electrons can be set to beforehand that matches the observed number that deflect in the detectors up or down at all measurement angles.
Keep in mind we're working with quantized values. Spin is always completely up or down, no matter what angle you measure it from. Measuring it at that angle locks it down, but tilting another detector in the chain after means you go back to a non-linear probability calculated by quantum mechanics (cosine of the angle in this case) to if it is up or down again at the new tilt and locks it down to that new tilt. There isn't any pre-determined value the electrons can be set to beforehand that matches the observed number that deflect in the detectors up or down at all measurement angles.