The whole idea is that the quantum particle can't have had the state you're measuring all along. If it did, then measuring a particular set of outcomes would be improbable. If you run an experiment millions of times, you have a choice in how you do the final measurement each time. What you find with quantum particles is that the measurements of the two different particles are more correlated than they should be able to be if they had determined an answer (state) in advance.

You can resolve this 3 ways:

1: you got extremely unlucky with your choice of measurement in each experiment lining up with the hidden/fixed state of each particle in such a way as to screw with your results. If you do the experiment millions of times, the probability of this happening randomly can be made arbitrarily small. So then, the universe must be colluding to give you a non uniform distribution of hidden states that perfectly mess with your currently chosen experiment

2: the particles transfer information to each other faster than the speed of light

3: there is no hidden state that the particle has that determines how it will be measured in any particular experiment

See https://www.quantamagazine.org/how-bells-theorem-proved-spooky-action-at-a-distance-is-real-20210720/ for a short explanation of what 'more correlated than expected' means