Seems like price is more predictive than beliefs right before an event, but beliefs are more predictive further out.
Possible that Metaculus algorithms can outperform pricing too though, if they combine pricing and beliefs, or correct beliefs well.
Related:
Use markets for what they are good at
Truly efficient markets can compensate for bias in participating populations
References:
http://journal.sjdm.org/18/18919/jdm18919.pdf?fbclid=IwAR08rKrq6DAGFmTp3K2p3hKWmpkRh5YAV6yl9aos1dDY7V9tmXHP3oar9Yg