I guess when you say "Are the particles connected through a wormhole or something like that", you are implying faster than light (FTL) signal.
The need for FTL signal, to explain entanglement behavior, actually arises from a confusion. It arises due to mixing of two types of correlations involved in entanglement.
- Anti-correlation (real time) - This means if you measure spin of an entangled particle in a certain direction, then the other particle will certainly have opposite spin if measured in the same angle. You can measure the two particles almost simultaneously to eliminate possibility of any signal at light speed, and the anti correlation is still found true. That makes some people think/say that there must be an FTL communication enabling this "real time" anti-correlation.
Actually, anti-correlation can be easily pre-determined like a pair of shoes and there would be no need of FTL signal. At the time of creation of entangled electrons, due to conservation of angular momentum, the wave function, imparted to the pair is capable of this behavior. So there is no problem with this case.
- Statistical Correlation - When spin of various entangled pairs is measured at two different angles A and B, then the correlation (same spin outcome) is given by QM predictions - square of cos((A-B)/2) times the number of pairs measured. Another scenario of statistical correlation is - If you keep measuring spin of an entangled electron in any specific angle, then 50% times, the spin will measure up and 50% times, it will measure down.
Statistical correlation is the one that perplexes people, because, it would require previous measurements to influence the new measurements. However, this would not require an FTL signal either. Because, to statistically influence the outcome of the new measurements, there would always be sufficient time for the light speed signal from previous measurements to travel to wherever it may be needed for the influence. (I am not implying that there is a light speed signal that influence the subsequent statistical outcome. All I am trying to say is that FTL signal is not really required)
Real time behavior would not require FTL signal because anti-correlation can be easily pre-determined.
Statistical behavior would not require FTL signal because it is statistical and there is plenty of time anyway.
Therefore FTL signal is not needed at all to describe entanglement behavior - unless you mix the two types of correlation.
However, based upon Bell's Inequality, some claim that in order to exhibit the statistical correlation, the anti-correlation can not be pre-determined. But that is a different issue for now.